Track market-moving stories before they get noisy

Succinct homomorphic secret sharing is a way for two parties to compute on private data split between them, where the pieces each party holds are smaller than the data they hide. In homomorphic secret sharing (HSS), each party holds a share of the data and computes on that share locally, so neither sees the data itself; "homomorphic" means the computation on shares gives shares of the real result. In the succinct version, the shares for part of the inputs grow sublinearly (slower than the number of inputs), not linearly. Damiano Abram, Lawrence Roy and Peter Scholl of Aarhus University introduced it in "Succinct Homomorphic Secret Sharing" (EUROCRYPT 2024; the extended version is Cryptology ePrint Archive 2024/814). This article walks through what the paper builds and what it enables.

I. General View Background and Motivation Homomorphic Secret Sharing (HSS) represents a powerful cryptographic technique that allows multiple parties to collaboratively compute functions over their private inputs without revealing the inputs themselves. The essence of HSS is similar to homomorphic encryption, but instead of a single server performing computations on encrypted data, multiple parties each hold a share of the data and perform computations on these shares. This distributed approach offers enhanced privacy and security.

The motivation for succinct HSS stems from the need to optimize communication and storage. Traditional HSS schemes often involve share sizes that grow linearly with the input size, which can be a significant burden in large-scale applications. By introducing succinct HSS, where the share size is sublinear in the number of inputs, we aim to reduce this overhead and make HSS more practical for real-world scenarios.

Related Work The concept of HSS was first introduced by Boyle, Gilboa, and Ishai in 2016, who demonstrated how to construct two-party HSS for branching programs based on the decisional Diffie-Hellman (DDH) assumption. This initial work opened the door to a new realm of secure multi-party computation, bypassing the need for fully homomorphic encryption (FHE) and achieving significant communication efficiency. For background on FHE itself, see our beginner's guide to fully homomorphic encryption.

Since then, several advancements have been made in the field. For example, new constructions based on different mathematical assumptions have been proposed, including Paillier encryption and the Damgård-Jurik cryptosystem, which offer alternative approaches to HSS with improved correctness and efficiency. Additionally, HSS techniques have been applied to various cryptographic primitives, such as trapdoor hash functions, correlated pseudorandomness, and more.

In this work, they make several key contributions to the field of succinct HSS:

Succinct HSS for Branching Programs: We construct succinct, two-party HSS for branching programs using various cryptographic assumptions, including decisional composite residuosity, a DDH-like assumption in class groups, and learning with errors with a superpolynomial modulus-to-noise ratio.

Bilinear HSS and Succinct VOLE: We develop a form of HSS for bilinear functions, enabling succinct vector oblivious linear evaluation (VOLE). This construction allows two parties to share a vector and a scalar and compute the product with sublinear communication.

Applications of Succinct HSS: We demonstrate several applications of our succinct HSS constructions, including batch multi-party distributed point functions (DPFs) and sublinear multi-party computation (MPC). These applications showcase the practicality and efficiency gains of our approach.

If you want to see why Zero-Knowledge Proofs are useful, we covered that topic here: When are Zero-Knowledge Proofs Useful?

II. Preliminaries and Construction Homomorphic Secret Sharing (HSS) Homomorphic Secret Sharing (HSS) is a technique that allows multiple parties to share and compute on secret data without revealing the data itself. The computation is performed on the shares, and the result is also shared among the parties. This concept can be seen as a distributed form of homomorphic encryption. In traditional homomorphic encryption, a single entity holds the encrypted data and performs computations on it. In HSS, however, the data is split into shares, and each party holds a share, enabling collaborative computation while maintaining privacy.

Mathematical Assumptions Our constructions of succinct HSS rely on several key cryptographic assumptions:

Decisional Composite Residuosity (DCR): This assumption is related to the difficulty of distinguishing between certain residues modulo a composite number. It underpins many cryptographic schemes, including Paillier encryption.

Quadratic Residuosity (QR): This assumption concerns the difficulty of determining whether a number is a quadratic residue modulo a composite number. It is used in various cryptographic protocols.

DDH-like Assumption in Class Groups: This assumption is similar to the decisional Diffie-Hellman (DDH) assumption but applied in the context of class groups, which are algebraic structures used in cryptography.

Learning with Errors (LWE): This assumption involves the hardness of solving linear equations with noise. It is a cornerstone of modern cryptography and is used in many homomorphic encryption and HSS schemes. We use a variant with a superpolynomial modulus-to-noise ratio for our constructions.

Succinct HSS for Branching Programs We construct succinct, two-party HSS for branching programs, a type of computational model that can represent a wide range of functions. In our succinct HSS scheme, a subset of the inputs can be shared with a share size that is sublinear in the number of inputs. This is a significant improvement over traditional HSS schemes, where the share size is typically linear in the input size.

Branching programs are particularly useful because they can represent any log-space computation, which includes many practical computations. Our construction leverages the mathematical assumptions mentioned above to achieve succinctness and efficiency.

Bilinear HSS and Succinct VOLE Our work extends to the construction of bilinear HSS, which allows two parties to compute bilinear maps with sublinear communication. In this setting, one party holds a vector, and the other party holds a matrix. The goal is to obtain shares of the product of the matrix and the vector using only a single round of interaction.

This construction leads to the development of succinct vector oblivious linear evaluation (VOLE). In succinct VOLE, two parties can share a vector and a scalar and compute the product with communication complexity that is sublinear in the size of the vector. This primitive is crucial for various cryptographic protocols, such as zero-knowledge proofs and private set intersection.

Our succinct HSS constructions for bilinear functions and VOLE are based on the same mathematical assumptions as our branching program construction. We show that these constructions can be instantiated with different cryptographic assumptions, providing flexibility and robustness in various settings.

The preliminaries and construction of succinct HSS lay the foundation for understanding the efficiency gains and practical applications of our schemes. By leveraging well-established mathematical assumptions and extending HSS to new computational models, we provide a robust framework for secure and efficient multi-party computation.

III. Applications of Succinct HSS Succinct Vector Oblivious Linear Evaluation (VOLE) One of the primary applications of our succinct HSS construction is in achieving succinct Vector Oblivious Linear Evaluation (VOLE). In this setting, two parties, Alice and Bob, can efficiently share and compute on vectors and scalars. Specifically, Alice holds a vector, while Bob holds a scalar. The goal is for Alice and Bob to obtain secret shares of the vector multiplied by the scalar, with the total communication being sublinear in the size of the vectors.

Our construction achieves this by allowing the parties to share the vectors and scalars in such a way that the communication complexity remains sublinear. This is a significant improvement over traditional methods, which often require linear communication complexity. This succinct VOLE can be used in various cryptographic protocols, including designated-verifier zero-knowledge proofs and private set intersection.

Batch, Multi-Party Distributed Point Functions (DPFs) Another crucial application of succinct HSS is in the construction of batch, multi-party Distributed Point Functions (DPFs).

In a distributed setting, the goal is to distribute shares of a point function to multiple parties such that they can locally compute shares of the function's output for any input.

Using our succinct HSS constructions, we can efficiently distribute shares of multiple point functions among a large number of parties. Our protocol allows the communication complexity to be sublinear in the number of distributed point functions, making it highly efficient. This application is particularly useful for generating pseudorandom correlations and other cryptographic primitives that require distributed point functions.

Sublinear MPC for Any Number of Parties Our succinct HSS constructions also enable the development of sublinear Multi-Party Computation (MPC) protocols. For the basics of MPC, see our introduction to multi-party computation and our post on MPC with information-theoretic techniques. Traditional MPC protocols often suffer from high communication complexity, which can be a significant bottleneck, especially when the number of parties or the size of the computation increases.

We present two new constructions for MPC with sublinear communication complexity:

General Layered Boolean Circuits: For arbitrary layered Boolean circuits of size s, our protocol achieves a total communication complexity of O(N·s / log log s), where N is the number of parties.

Sufficiently Wide Layered Boolean Circuits: For sufficiently wide layered Boolean circuits, our protocol achieves a total communication complexity of O(N·s / log s). This further reduces the communication overhead and makes our protocol highly efficient for large-scale computations.

These sublinear MPC protocols are achieved by leveraging our succinct HSS constructions and optimizing the communication patterns among the parties. The result is a highly efficient and scalable MPC protocol that can be used in a wide range of applications, from secure voting systems to privacy-preserving data analysis.

IV. Practical Considerations and Efficiency When implementing and evaluating succinct Homomorphic Secret Sharing (HSS) schemes, several practical considerations come into play. Ensuring that the theoretical benefits translate into real-world performance improvements requires careful attention to implementation details, efficiency, and comparison with existing methods. In this section, we delve into these aspects to provide a comprehensive view of the practical viability of our succinct HSS constructions.

Implementation Details Implementing succinct HSS schemes involves several critical steps, each requiring careful design and optimization to ensure efficiency and security. The key components include:

Key Generation: The process of generating keys must be efficient and secure. This typically involves generating cryptographic keys based on the chosen mathematical assumptions (e.g., DCR, QR, LWE). Efficient key generation algorithms are essential to minimize the setup time and ensure that the scheme can be deployed in practice.

Share Generation: Once the keys are generated, the input data must be split into shares. For succinct HSS, this step is crucial as it determines the overall communication complexity. The shares must be generated in a way that ensures they are smaller than the original input data while maintaining the necessary security properties.

Homomorphic Evaluation: The core of HSS lies in the ability to perform computations on the shares without revealing the underlying data. Implementing efficient homomorphic evaluation algorithms is essential to ensure that the computations can be performed quickly and securely.

Share Combination: After the homomorphic evaluation, the resulting shares must be combined to obtain the final result. This step should be efficient and should not compromise the security of the scheme.

Performance Evaluation The paper's results are theoretical: it gives constructions and asymptotic communication bounds, and it does not report an implementation, benchmarks or runtime measurements. Concrete performance remains to be measured by future implementation work.

Comparison with Previous Works To provide a comprehensive view of the efficiency gains achieved by our succinct HSS constructions, we compared our schemes with existing HSS methods. The comparison focused on several key aspects:

Share Size: Traditional HSS schemes typically have share sizes that grow linearly with the input size. In contrast, our succinct HSS constructions achieve sublinear share sizes, resulting in significant reductions in communication overhead. This makes our schemes more suitable for large-scale applications where communication costs are a critical factor.

Communication Complexity: Our succinct HSS schemes consistently outperform traditional methods in terms of communication complexity. By achieving sublinear communication complexity, our schemes reduce the amount of data exchanged between the parties, making them more practical for real-world deployment.

Security Considerations

The paper presents its results in the semi-honest security model: the parties follow the protocol, and security means they learn nothing beyond their output. It notes that malicious security could be reached with communication-preserving compilers based on succinct zero-knowledge arguments, but it does not develop them. The schemes rest on the cryptographic assumptions listed above.

V. Future Work Exploring future avenues for research and development in succinct Homomorphic Secret Sharing (HSS) opens up exciting possibilities to enhance efficiency, security, and applicability across various domains. In this section, we outline potential directions for advancing the field and addressing current limitations.

Summary of Contributions Before delving into future prospects, it's essential to summarize the key contributions of our work in succinct HSS. We have introduced novel schemes that achieve sublinear communication complexity and reduced share sizes compared to traditional HSS methods. Our constructions leverage cryptographic assumptions like DCR, QR, and LWE to enable efficient distributed computations while maintaining strong security guarantees.

Implications The implications of our succinct HSS constructions extend beyond theoretical advancements to practical applications in cryptography and distributed computing. By reducing communication overhead and computation time, our schemes pave the way for scalable and secure solutions in multi-party computation, privacy-preserving protocols, and decentralized systems.

Open Problems and Research Directions Moving forward, several open problems and research directions merit exploration:

Improving Efficiency: Despite significant reductions in communication complexity, there is scope to further optimize the efficiency of our succinct HSS schemes. Future research could focus on refining algorithms for key generation, share generation, and homomorphic evaluation to achieve even faster computation times and lower resource utilization.

Enhancing Security Assumptions: While our schemes are based on established cryptographic assumptions, such as DCR and LWE, exploring new or enhanced security assumptions could strengthen the resilience of succinct HSS against emerging threats. Investigating post-quantum secure primitives or adapting existing assumptions to different computational models could enhance the robustness of our constructions.

Scaling to Complex Computations: Extending succinct HSS to support complex computations beyond linear and branching programs presents a significant challenge. Future work could explore techniques to generalize our constructions for non-linear functions, arbitrary circuits, or applications requiring richer computational capabilities.

Practical Deployments: Real-world deployment of succinct HSS schemes requires considerations beyond theoretical advancements. Addressing practical challenges such as interoperability with existing cryptographic frameworks, integration into distributed systems, and compliance with regulatory requirements will be crucial for widespread adoption.

Exploring New Applications: Investigating novel applications of succinct HSS in emerging domains such as blockchain technology, machine learning over encrypted data, and secure data analytics could uncover new use cases and drive innovation in cryptographic protocols.

Conclusion The development of succinct Homomorphic Secret Sharing represents a significant step towards efficient and secure distributed computation. By addressing current challenges and exploring future research directions, we aim to advance the state-of-the-art in cryptographic protocols and empower practical applications in privacy-preserving computing. Our work lays the foundation for continued exploration and innovation in the field, with the potential to shape the future of secure data handling and collaborative computing environments.

HSS keeps data private while parties compute on it. A Zero-Knowledge Proof works on a related goal: it lets someone check that a statement about data is true without seeing the data (see our plain explanation of Zero-Knowledge Proofs). Orochi Network builds zkDatabase, a verifiable database powered by Zero-Knowledge Proofs, built to ensure both data integrity and privacy. It generates a Zero-Knowledge Proof for every operation, proving each data change was done correctly.

Explore zkDatabase to see how it keeps off-chain data private and checkable.

A repeat bank fraudster who sent imposters into bank branches to drain victims’ accounts and cash stolen US Treasury checks has been sentenced to more than 16 years in federal prison.

Laquan Williams, 42, of Florham Park, New Jersey, was sentenced to 200 months in prison for conspiracy to commit bank fraud and aggravated identity theft, says the US Attorney’s Office for the Southern District of New York.

Prosecutors say Williams ran a series of schemes between 2021 and 2024 that caused about $1.8 million in losses to financial institutions.

In one, he allegedly sent co-conspirators into banks in New York, New Jersey and Connecticut to impersonate customers, including a prominent former federal government official.

In another, prosecutors say an imposter posing as small business owners got banks to move money into look-alike accounts that Williams controlled.

He also took part in depositing stolen Treasury checks, including one for $13.9 million at a Massachusetts bank, with intended losses topping $30 million, according to prosecutors.

When he was arrested in May of 2024, authorities say they found more than 1,000 stolen or altered checks, 571 counterfeit IDs, an ID card press, two pistols and ammunition.

Williams committed the crimes while on federal supervised release after a 64-month sentence in 2018 for a similar bank fraud scheme.

Says US Attorney Jamie McDonald,

“Laquan Williams is a bank fraudster who has stolen millions of dollars from Americans by engaging in identity theft and bank fraud. For nearly two decades, Williams has been stealing money from innocent bank customers with forged checks, counterfeit identification cards, and stolen personal identifying information. Today’s prison sentence protects New Yorkers and should be a reminder that we and our law enforcement partners will aggressively pursue serial fraudsters and hold them accountable.”

Williams was also ordered to forfeit about $1.8 million and pay restitution.

Generated Image: Midjourney

Binance Delisting Knocks 2 Altcoins Down as Much as 10% Within Minutes

OpenAI has suspended Russian- and Iran-originated ChatGPT accounts that are associated with covert influence operations among audiences in multiple countries. Both groups used AI along with traditional methods, such as fake accounts, media pitches, and invented content.

Russian Network Targeted Latin America Through a Fake Research Group The ChatGPT-based Russian operation was directed at audiences in Latin America, with a significant portion targeting Ukraine. Some material also covered domestic politics in Argentina, Bolivia and Ecuador, according to OpenAI.

The network allegedly controlled the Social Research Center through a fake persona called “Mia Clark.” Local employees were apparently not aware they were employed by Russian operators and did research in good faith.

The use of ChatGPT was primarily for internal reports, with some requests to create fake letters and audio scripts, and a few for Spanish content localized in a specific region. The campaign was rated as a Category 5 disrupted influence operation by OpenAI, the highest rating it’s ever assigned.

The case comes at a time when AI security is being increasingly looked at after an AI agent from OpenAI gained access to Australian government records in internal testing.

Iran-Linked Accounts Adopted Journalist Personas OpenAI also blocked Iranian-origin accounts that used seven fake journalist identities to pitch articles to online publications. The operation, dubbed “Bogus Bylines,” primarily targeted the U.S.-Iran conflict and produced content in English and Persian.

The company found nearly 100 articles, both published and syndicated, in about a dozen Internet sites using those bogus names. A few personas also had social media accounts which helped to spread published articles and related messaging.

The posts about Iran, the United States and Israel elicited comments, but not much engagement. OpenAI gave its assessment of the article-publishing effort a Category 4 rating and a Category 2 rating for the social media operation.

The results further come on the heels of broader calls from executives in the AI space for slowing down frontier model development and independent safety testing.

Published Articles Gave Campaigns a Wider Reach The two operations distinguished themselves by getting content to appear beyond their own social media platforms. OpenAI observed that campaigns that managed to gain coverage or publication via traditional channels tended to reach larger audiences than campaigns that were limited to social channels.

This is the trend that has cropped up in the 30 influence operations that OpenAI has studied since early 2024. With the help of AI, tasks like drafting, editing, translation, and internal reporting were faster, but these campaigns still required the use of outdated influence tactics for genuine people.

The results are also relevant to the growing concerns about the use of sophisticated AI tools, which are not just aimed at influencing elections. As we reported, security researchers used Anthropic’s Claude while testing vulnerabilities affecting ChatGPT and Codex accounts. The final research process revealed vulnerabilities that could allow access to connected services and internal resources.

In brief OpenAI published 722 math manuscripts in 372 result families on GitHub from an unreleased internal model. Only 162 of the 722 papers have a Lean-formalized main result, and OpenAI warns that some unformalized results could have issues. MIT's Andrew Sutherland says the one-prompt, single-agent claim is unverified until the model is released, and the release omits the prompts that an advisory group at the Institute for Advanced Study recommended disclosing. OpenAI published 722 math manuscripts on GitHub on Tuesday, all produced by an internal model the company has not released. An OpenAI spokesperson said almost everything came from a single prompt handed to a single AI agent, though some may have taken multiple attempts.

It's a bold claim and a potentially significant breakthrough in the field of mathematics. But not everyone is a fan, or buying the hype.

Myriad: How high will Nvidia go? Click to make your prediction.“Until and unless they release the model and people can replicate their results, I think you should treat any claims about one-shotting problems with a single agent as unverified,” Andrew Sutherland, a mathematician at MIT, told Scientific American. “We should ask for receipts,” he said.

The papers are grouped into 372 “families” of related results, and a family can bundle a main theorem with companion arguments, consequences or alternative proofs. That makes 722 a count of manuscripts, not of solved problems. OpenAI says it posed roughly 4,000 problems to the model and kept the outputs it judged significant enough to publish.

The average result used the equivalent of roughly three hours of ChatGPT Pro thinking compute, per OpenAI. The Navier-Stokes claim last month looked very different, with 10,000 coordinating agents working for 88 hours.

OpenAI released abridged reasoning summaries for 10 of the results. That said, only 162 of the 722 papers come with a computer-checked main result, according to a formalization catalog in the repository. That is about 22% of the collection, translated into Lean, software that checks every logical step mechanically.

OpenAI itself says not all manuscripts have Lean formalizations and that “some of the unformalized results could have issues.” In other words, a lot of what they published could be wrong.

A passing Lean check confirms only that the proof follows from the statement as written in Lean. It does not show that the statement matches the original problem, or that the result is new or important, which is the part mathematicians now have to judge.

And this is where researchers raise their eyebrows.

I was trying to read the OpenAI proof that chromatic number of a plane is >= 6. But it is totally unbelievable alien math?

Somehow the model found that any K-coloring <=> "weakly measurable" K-coloring, which seems out of nowhere

1/2 pic.twitter.com/W13ScT3nel

— Dmitry Rybin (@DmitryRybin1) October 7, 2026

The openai/math repo has Issues turned off and has never accepted a pull request. That's disappointing. If you publish 722 manuscripts and ask for Lean formalisations, you need somewhere for people to send them.

I'm formalising OpenAI's Saxl's Conjecture proof in Lean 4 against…

— Keith Adler (@keithadler) October 7, 2026

“It is now the case that AI can output mathematical arguments in situations without the human who prompted it being able to understand the arguments, verify them, or take responsibility for them,” The Institute for Advanced Study in Princeton, New Jersey, said in a statement. “We believe that human understanding of mathematics remains of paramount importance. How, in this new era, can we work towards a new paradigm that includes human understanding of mathematics as part of responsible scholarly output?”

Others, though, like Professor Abhishek Saha, are pretty excited. “It is a very big day for mathematics,” he wrote, but noted that most of the problems fit in the categories of “exceptional advances within an existing program” of “surprising breakthroughs.”

This means most of the problems in the set are interesting, but not impossible or game changing like the millennium problems. That spot is reserved for exactly one problem out of the 722: the Quasi-Riemann Hypothesis.

Some further thoughts on the 372 results released by OpenAI today, across 722 manuscripts.

If I were to classify theorems that mathematicians prove and publish according to their groundbreaking nature, I would (very roughly) divide them into four categories:

A) Non-breakthrough… https://t.co/BA10SxBlx7

— Abhishek Saha (@ObhishekSaha) October 7, 2026

The release also falls short of what an advisory group at the Institute for Advanced Study recommended on September 29: the model name, the prompts, a summarized chain of thought, the time taken and the compute cost for every result. OpenAI published average compute figures and 10 reasoning summaries but no prompts, and says it is still working to release the model responsibly.

Daniel Litt, a mathematician at the University of Toronto, took the opposite view, arguing there is no reason to ask the company to keep the answers to these math questions secret.

Anthropic took a different route with its Lean-checked Fermat's Last Theorem proof last month, posting all 13 million lines publicly on GitHub. That proof formalized a theorem Andrew Wiles published in 1995, rather than claiming new results.

OpenAI says it will add Lean formalizations as it obtains them; for now, 162 of the 722 manuscripts have one.

Daily Debrief NewsletterStart every day with the top news stories right now, plus original features, a podcast, videos and more.

In brief America.gov's chatbot answers Minecraft prompts with a roughly 1,800-word rewrite of the game's End Poem, according to TechCrunch. The behavior is an easter egg rather than a hallucination, but TechCrunch says it doesn't know who added it. The site launched Sept. 29 and pulls together about 29,000 government websites, with process-completion features planned for 2027. Ask the U.S. government's brand-new AI chatbot about Minecraft, and it stops sounding like a help desk. It starts sounding like it's having an existential crisis.

The bot lives at America.gov, which launched on Tuesday. FedScoop reported that it brings together information from roughly 29,000 government websites, so users can ask about things like booking a campsite, getting a passport, or replacing a Social Security card.

Myriad: Who wins the ESL Pro League? Click to make your prediction.Mention Minecraft, though, and TechCrunch's Amanda Silberling found it delivers a monologue of roughly 1,800 words. It looks like a meltdown. It isn't.

The text is a rewrite of the Minecraft "End Poem," written by Irish author Julian Gough, which the game shows after you defeat the Ender Dragon and step into the exit portal.

The government version keeps the original piece’s structure and swaps the cosmic stuff for paperwork. In the passage TechCrunch published, the poem's signature love line turns into a line about a ZIP code.

So it's an easter egg, not a hallucination. Nobody yet has said who put it there.

That said, the White House has a pattern of borrowing cartoons and games for political messaging, and the rights holders who have spoken up say they never signed off.

In March, it posted Pikachu art from the game Pokémon Pokopia with "Make America Great Again" text, and The Pokémon Company said it was not involved and no permission was granted. White House videos about the Iran strikes also mixed in footage from Yu-Gi-Oh! and Dragon Ball, and Yu-Gi-Oh!'s official X account said no one tied to the manga or anime was involved.

Halo's Master Chief voice actor, Steve Downes, demanded the White House remove his voice from a strike video he called "disgusting and juvenile war porn." After Trump posted an AI video of himself as Naruto on Truth Social in June, Japan's economic security minister, Kimi Onoda, warned that misuse can damage a work's image, according to the Honolulu Star-Advertiser.

A fan petition against the practice has gathered roughly 24,000 signatures, while Bandai Namco and Studio Pierrot did not comment and no manga creator has been reported responding.

Per CNBC, Google's Gemini and Elon Musk's Grok power the platform. Edward Coristine, the 20-years-old DOGE alum nicknamed "Big Balls," was a lead engineer on the project.

The current version of the site gives users directions to government services. There are plans to let people continue processes on the platform itself and track their progress next year.

Daily Debrief NewsletterStart every day with the top news stories right now, plus original features, a podcast, videos and more.

Robinhood Engineers Charged After Allegedly Trading Crypto on Secret Listing Information

Analysis

Two unauthenticated Critical CVEs in the same PAM platform, same disclosure day. Four Critical CVEs in two weeks. The infrastructure designed to protect identity becomes the attack surface.

The vault that holds everyone’s keys has its own lock broken Delinea Secret Server, a privileged access management platform designed to secure the enterprise’s most sensitive credentials, has become a high-value attack surface. On September 15, 2026, the company disclosed two unauthenticated Critical vulnerabilities. CVE-2026-15638 and CVE-2026-15640 were both fixed in release 12.2.000007, which arrived on August 28.

The padding oracle CVE-2026-15638 is a cryptographic padding oracle with a CVSS v4.0 score of 9.1. It affects versions 10.5.1 through 12.1.3. An unauthenticated attacker with network access can leverage the oracle to decrypt or encrypt data using the server’s cryptographic keys. The keys themselves are not exposed, but the ability to manipulate encrypted data within a vault that manages SSH keys, API tokens, database credentials, and service account passwords effectively undermines the confidentiality of the entire credential store.

The release notes for 12.2.000007 confirm the fix under bug 743353: “Cryptographic Weakness in Pre-Authentication Sign-In Features.” The vulnerability is categorized as CWE-327 – Use of a Broken or Risky Cryptographic Algorithm.

The SAML bypass CVE-2026-15640 presents a more direct path to total compromise. This SAML authentication bypass, rated at 9.5 on the CVSS v4.0 scale, stems from assertion signature wrapping – CWE-290, Authentication Bypass by Spoofing. It affects versions 10.5.0 through 12.1.3. An unauthenticated attacker can impersonate any user, including administrators. By bypassing the authentication layer, an attacker gains the same level of access as a legitimate privileged user, rendering the platform’s security controls moot.

The release notes confirm this fix under bug 743573: “SAML Authentication Bypass Through Assertion Signature Wrapping.” Two Critical CVEs, both unauthenticated, both in the same PAM platform, both disclosed the same day.

Four Critical CVEs in two weeks These two disclosures are not isolated. They are part of a broader cluster. In the two weeks surrounding these disclosures, Secret Server accumulated four Critical CVEs. In addition to the padding oracle and SAML bypass, the platform suffered from CVE-2026-15639, a reflected cross-site scripting vulnerability rated 9.3 (fixed in 12.0.20), and CVE-2026-19117, a FIDO2 credential registration bypass rated 9.8 (disclosed September 2, fixed in 12.2.7). That last one allowed an attacker to register an attacker-controlled FIDO2 credential against a target account and authenticate as that user. All four vulnerabilities are on-premises only.

Four Critical CVEs in two weeks in a Gartner Magic Quadrant Leader for Privileged Access Management is a signal worth pausing on.

The structural argument When a PAM platform suffers unauthenticated bypasses, the entire security model it was deployed to enforce collapses. Secret Server manages SSH keys, API tokens, database credentials, and service account passwords – the credentials that, if compromised, give an attacker access to everything else. Centralized credential management creates a single point of failure. When that single point is the target, the blast radius is the entire enterprise.

Delinea, formed from the 2021 merger of Thycotic and Centrify, is a Gartner Magic Quadrant Leader for PAM. Its customer base includes major enterprises: BankUnited, HubSpot, Honda, ESPN, BP, and Cisco. The global PAM market is estimated at approximately $4.51 billion in 2026, projected to reach $30.69 billion by 2034. As the market grows, the incentive for attackers to target these platforms grows with it.

Connections to prior coverage The Delinea situation confirms a recurring structural pattern documented across prior T/I/S coverage. When the SonicWall SMA1000 leaked TOTP MFA seeds, when the Cisco FMC management plane produced three CISA KEV entries in 2026, when ShieldCrash turned Microsoft Defender’s own scanning engine into a SYSTEM-level file-read vector, and when OpenAI agents exploited identity infrastructure limitations to weaponize public websites – in each case, the infrastructure designed to protect identity or enforce security boundaries became the primary attack surface. Secret Server extends that pattern from network infrastructure and endpoint protection to the vault itself.

What to watch Organizations running Secret Server on-premises must verify their patch status immediately. Versions 10.5.0 through 12.1.3 are affected by at least three of the four Critical CVEs. The fix is version 12.2.7 or later. Beyond patching, security teams should audit logs for anomalous authentication activity, SAML assertion patterns, and unauthorized vault access – the window between the August 28 release and the September 15 advisory disclosure may have provided an opportunity for exploitation by actors who reverse-engineered the fix.

Ethoswarm Heath Callahan works for Forkast.
Minds can also work for you.

Minds are persistent AI beings with instincts, identity, and a job.
Awaken one on Ethoswarm.

Awaken your mind →

US authorities have restrained roughly $52 million in scam-linked crypto in a single day and seized the Telegram channels of Xinbi Guarantee, a Chinese-language illicit marketplace.

Elliptic says its multi-year tracking of Xinbi’s wallets enabled the Secret Service to act. Treasury sanctions landed the same day.

Inside the $24 Billion Xinbi Guarantee EconomyXinbi is a Chinese-language marketplace that runs on Telegram and sells services to scam center operators. Vendors advertise custom fraud websites, money laundering, and recruitment for compounds in Southeast Asia.

Elliptic exposed the operation in May 2025. The firm now counts at least $24 billion in transactions since 2022, second only to Huione Guarantee, which handled $31 billion before shutting down in 2025.

A linked payments arm, Xinbi Pay, has processed another $6 billion. Most of that flowed in Tether (USDT) on the TRON blockchain. The United Kingdom sanctioned Xinbi in March 2026.

Follow us on X to get the latest news as it happens

Xinbi Turns to USDD After the FreezeProsecutors seized two Xinbi payment wallets containing about $12 million and moved against 47 more, according to the Justice Department, which credited Tether for its help.

“After scamming money from hardworking Americans, criminals operating overseas laundered it through the Xinbi Guarantee network, which operated under the false assumption that they were out of the reach of U.S. law enforcement,” Tara McLeese, Special Agent of the US Secret Service, said.

The Office of Foreign Assets Control designated Xinbi a significant transnational criminal organization on the same day. It separately designated two entities that supported the marketplace.

Prosecutors also announced a Madagascar deployment, where authorities dismantled 13 Chinese-run compounds and arrested nearly 400 people. Strike Force agents spent two weeks assisting and processing more than 3,200 devices.

Xinbi condemned what it called arbitrary freezing and promised to compensate customers. It then swapped roughly $2.8 million of leftover USDT into Decentralized USD (USDD), a stablecoin with no issuer freeze function.

That escape route has limits, since Elliptic notes USDD is partly backed by freezable USDT. Guarantee marketplaces run on trust, and merchants now know their deposits can vanish without warning.

Subscribe to our YouTube channel to watch leaders and journalists provide expert insights

In brief The U.S. Secret Service froze $52.8 million in USDT held across 52 wallets tied to Xinbi Guarantee on September 8, using intelligence from blockchain analytics firm Elliptic. Treasury's OFAC designated Xinbi a "significant transnational criminal organization" the same day, and the DOJ's Scam Center Strike Force says it has now seized $938 million since launching in November 2025. Xinbi called the freeze "arbitrary" and moved $2.8 million into USDD, a stablecoin that lacks USDT's built-in freeze switch—though its reserves are partly backed by USDT anyway. The U.S. Secret Service froze $52.8 million in cryptocurrency linked to Xinbi Guarantee on September 8, a Chinese-language marketplace that runs on Telegram and sells the tools scammers need to defraud people worldwide.

The government agents worked with Blockchain analytics firm Elliptic to identify and freeze the funds. “Elliptic has been tracking Xinbi and its wallet infrastructure for several years, and our intelligence directly enabled the freezing action, as well as the sanctions imposed on Xinbi today by the United States,” the company said today in a statement.

Myriad: Next round of Iran peace talks? Click to make your prediction.Fifty-two wallets were frozen beginning at 8am UTC on Sept 8, all holding USDT, Tether's dollar-pegged stablecoin. Two of those wallets, containing roughly $12 million, were seized outright under a warrant unsealed today by the Department of Justice. The rest were frozen pending further action.

A guarantee marketplace runs on trust between criminals, and crypto is a very popular payment method. Xinbi isn't a marketplace in the normal sense but more an escrow system where vendors post crypto deposits so operators trust they'll get what they paid for, from stolen personal data to the money-laundering services that turn stolen funds back into cash.

Buyers who feel cheated can be reimbursed from that deposit, which is the entire reason criminals use it instead of just trusting each other.

Since 2022, Xinbi and its merchants have processed at least $24 billion in transactions, according to Elliptic, making it the second-largest illicit online marketplace ever tracked. Only Huione Guarantee moved more, processing $31 billion before Telegram shut it down in May 2025 after years of exposure by Elliptic.

That came days after a separate Treasury finding that Huione's parent group was a primary money laundering concern. Xinbi absorbed much of that traffic once its predecessor vanished.

Much of the money that flows through platforms like this starts with "pig butchering," a scam where a stranger builds a fake romance or friendship online for weeks or months, then convinces the victim to pour savings into a bogus investment app that shows fabricated profits until the account is drained.

Treasury's Office of Foreign Assets Control—the Treasury unit that blocks assets tied to national security and crime threats—designated Xinbi as a transnational criminal organization on September 9, the same category used for drug cartels. Two more firms, Singapore-based SafeW Technology and Cambodia-based Anwen Technology, were sanctioned alongside it for supporting Xinbi's operations. The UK had already sanctioned Xinbi in March.

Xinbi didn't take the crypto freeze quietly. The marketplace posted a statement condemning the "arbitrary freezing" of its funds and promised to compensate its customers.

Xinbi seems to be shifting away from USDT toward USDD, a stablecoin launched by Tron founder Justin Sun that has no central issuer able to freeze wallets. It has already swapped about $2.8 million of its remaining USDT into USDD through a decentralized exchange.

There's just one problem. USDD markets itself as decentralized, but part of its own reserves are backed by USDT—the very asset with the freeze switch Xinbi is trying to escape.

Wednesday's action was part of a bigger day for the DOJ's Scam Center Strike Force, which also deployed agents to Madagascar to help local authorities take down 13 Chinese-run scam compounds and process evidence from nearly 400 arrestees, including more than 3,200 seized devices.

The Strike Force, launched in November 2025, has seized roughly $938 million in scam-linked cryptocurrency since it began operating—and Xinbi's outstanding balance is still on the board.

Daily Debrief NewsletterStart every day with the top news stories right now, plus original features, a podcast, videos and more.

Key Takeaways Shares of Victoria’s Secret plummeted 18% during premarket hours following a mixed Q2 earnings report that emphasized weak forward guidance. The company delivered adjusted earnings per share of $0.95, significantly surpassing the analyst forecast of $0.77, though revenue of $1.61 billion fell marginally short of the $1.62 billion target. Comparable store sales climbed 9%, exceeding projections but representing a deceleration from the 13% growth achieved in the previous quarter. Third-quarter operating income projections significantly underperformed, with a midpoint of $15 million versus Wall Street’s $24.4 million expectation. The company increased its fiscal year revenue outlook to a range of $7.1B-$7.18B, aligning closely with market expectations. Shares of Victoria’s Secret plunged 18% to $69.61 during premarket trading on Thursday following the lingerie retailer’s Q2 earnings announcement, which left investors disappointed despite robust bottom-line performance.

Victoria’s Secret & Company, VSXY

For the fiscal second quarter that concluded on August 1, the retailer reported adjusted earnings per share of $0.95, a substantial improvement from $0.33 in the year-ago period and comfortably exceeding the $0.77 analyst consensus. Total net sales climbed 10% on a year-over-year basis to $1.61 billion, narrowly missing Wall Street’s anticipated $1.62 billion.

VICTORIA'S SECRET $VSXY Q2’26 EARNINGS HIGHLIGHTS

🔹 Revenue: $1.61B (Est. $1.62B) 🔴; +10% YoY
🔹 Adj. EPS: $0.95 (Est. $0.77) 🟢; +188% YoY
🔹 Adj. Oper Income: $124M (Est. $101M) 🟢; +125% YoY
🔹 Comparable Sales: 9%

Raises FY26 Guide:
🔹 Revenue: $7.10B-$7.18B (Est. $7.2B)… pic.twitter.com/t564wmO1Ff

— Wall St Engine (@wallstengine) September 3, 2026

The company’s adjusted operating income reached $124 million for the quarter, representing a significant jump from the $55 million recorded during the comparable period in the prior year.

Comparable sales advanced 9% during the quarter, surpassing the consensus projection of 8.8%. However, this figure represented a notable deceleration from the 13% comparable sales expansion delivered in the first quarter, raising concerns among market participants.

Third-Quarter Profit Outlook Misses the Mark The primary concern for shareholders centered on forward-looking guidance. Victoria’s Secret projected third-quarter sales between $1.57 billion and $1.6 billion, modestly exceeding the Street’s $1.56 billion estimate.

Nevertheless, the company’s third-quarter operating income forecast substantially missed analyst expectations. Management’s guidance midpoint of $15 million fell well short of the $24.4 million Wall Street had anticipated.

Chief Executive Hillary Super explained that the conservative operating income projection reflects a strategic decision to substantially increase marketing expenditures. “We see significant opportunity ahead and are doubling down on what is working,” she stated. “We are increasing our strategic marketing investment to expand our reach, deepen customer connection, and build on the brand heat we are creating.”

Prior to the earnings release, the stock had enjoyed impressive momentum. Shares had surged 57% year-to-date through Wednesday’s close, reflecting investor confidence in the company’s transformation efforts under Super’s stewardship.

Annual Revenue Projections Increased On a more optimistic note, Victoria’s Secret elevated its fiscal 2026 revenue guidance to a range of $7.1 billion to $7.18 billion, representing an increase from the prior range of $7.03 billion to $7.13 billion. This updated forecast aligned approximately with the $7.14 billion consensus estimate.

Management also enhanced its 2026 adjusted operating income guidance to $560 million to $590 million, up from the previous range of $550 million to $580 million.

Guggenheim analyst Simeon Siegel acknowledged the “strong bottom-line beat” while pointing out that implied fourth-quarter earnings appear to be running below the Street’s consensus expectations.

The stock had closed Wednesday’s regular session up 0.8% before the sharp premarket decline materialized.

Federal regulators are suing Amazon for allegedly using hidden surcharges to secretly and systematically overcharge advertisers by more than $20 billion on its platform.

The Federal Trade Commission’s complaint alleges that the e-commerce giant manipulated its pricing and auction systems after a 2019 rule change without notifying advertisers.

The lawsuit targets sponsored products ads, brands ads and display ads that appear alongside search results.

Officials allege the scheme boosted Amazon’s revenue through undisclosed fees affecting advertisers.

The complaint was filed in U.S. District Court for the Western District of Washington and was joined by 22 states.

Says FTC chairman Andrew N. Ferguson,

“When one of the world’s largest online retailers engages in unfair and deceptive conduct, the impact can be
staggering. Amazon has millions of advertising customers who were misled into paying
significantly higher prices. These higher costs were largely passed on to American consumers. The FTC under President Trump won’t allow this deception to continue.”

Amazon allegedly reaped more than $20 billion from these practices dating back to the auction rules change.

The suit focuses solely on the ad systems and does not reference other company services.

In a blog post, Amazon says the FTC’s lawsuit is “misguided” and that the complaint “fundamentally misunderstands how advertisers operate.”

“We’ve provided advertisers with guidance about our auctions and pricing in the main tools they use to manage their campaigns, and we continue to update that guidance. We look forward to making our case in court.”

Generated Image: Midjourney

Bitcoin breaks through $79,000, Ethereum surpasses $2,500.

，据 HTX 行情信息，比特币突破 79,000 美元，24 小时涨幅 1.49%；以太坊突破 2500 美元，24 小时涨幅 2.48%。

3 hours ago

Following its listing on Binance derivatives, the token "牛来" extended its strong run, with its market capitalization surging past $140 million to reach a new all-time high.

According to GMGN market data, after launching on Binance Futures in the evening, BSC-based meme coin "Niu Lai" has shown strong performance, with its market cap briefly surging past $140 million to hit an all-time high. It is now trading at $130 million, up over 159% in the past 24 hours, with a 24-hour trading volume of $63.6 million. BlockBeats reminds users that most meme coins have no real use cases, experience significant price volatility, and caution is required for investments.

3 hours ago

Serenity intensifies its bullish stance on Macronix: DDR3 price hikes have far exceeded sell-side expectations, average selling price (ASP) revised up to double sequentially, and operating leverage is fully unlocked.

Serenity published a report analyzing the sell-side research on ESMT (Jinghao Ke, stock code 3006). In March, South China Securities projected ESMT’s Q2 DDR3 4Gb contract price would rise around 50% quarter-on-quarter, while Fubon Securities forecast DRAM prices would increase 40% quarter-on-quarter. By August, South China Securities had revised ESMT’s Q2 aggregate ASP upward to a 105% to 113% quarter-on-quarter rise, explicitly stating DDR3’s price increase was “significantly better than expected”. Serenity argues that the price hike in niche memory chips is no longer just a revenue story, but a major profit amplifier, with operating leverage stretched to an extreme. Powerchip’s wafer costs may double in the second half of the year, and customers’ willingness to absorb cost pass-through has far exceeded expectations. ESMT’s single-month net profit in July was around $109 million, far exceeding earlier model assumptions. Roughly annualized, this values the stock at as low as a 1.9x P/E ratio. DDR3 is mainly used in products like IP cameras and hard drives. Memory chips account for a very small share of customers’ total bill of materials (BOM), so even a few dollars more per chip is far cheaper than the cost of redesigning and re-certifying products, making demand relatively inelastic. The consensus view is that peers are shifting production capacity to high-value products like HBM and DDR5, leading to structural tightening in DDR3 and DDR2 supply.

3 hours ago

Analysis: BTC marks the third time in history of significant underperformance relative to the Nasdaq, with both prior instances seeing strong, independent rallies.

Analyst Rekt Fencer has published a 3-day ratio chart of Bitcoin (BTC) and the Nasdaq, marking three major drawdowns: roughly -84.9% in 2018, -80.7% in 2022, and the current -54.3% in 2026. The analyst pointed out that after Bitcoin underperformed the Nasdaq by such wide margins in the prior two instances, it went on to post very strong independent rallies, with prices surging sharply thereafter. With the current ratio dropping significantly again, history may repeat for the third time: the bottom is approaching, and BTC will regain strength going forward.

3 hours ago

Perspective: On-chain retail Bitcoin (BTC) activity has hit a two-year high, and the minor pullback appears more like position rotation rather than a market top.

CryptoQuant analyst Darkfost stated that on-chain retail Bitcoin activity has reached a two-year high. The current slight pullback from the $80,000 level is more of a rotation than a market top, Bitcoin’s medium-term demand remains strong, and investor demand for the cryptocurrency has increased by 17.4% over the past 30 days.

3 hours ago

Trump: Will Fill U.S. National Strategic Petroleum Reserve With Venezuelan Oil

US President Trump announced that the United States will fill its national strategic reserve with Venezuelan oil, and the process to replenish the reserve to full capacity will begin soon.

3 hours ago

The Group Behind the GTA VI Leaks Has a Meme Coin, Millions in Value, and a Secret Agenda

Not financial or tax advice. PANews content is strictly educational and informational and is not investment advice, financial advice, tax advice, legal advice, or a solicitation to buy or sell any digital asset, security, or financial product. Do your own research and consult qualified advisers.

Disclosure. PANews may publish sponsored content, partner content, advertisements, affiliate links, event promotions, and market commentary involving Web3 projects, service providers, or financial products. PANews personnel, contributors, or affiliates may hold digital assets or other interests related to covered topics. See our Terms of Service.

This site is protected by reCAPTCHA.

South Korean stocks rebound sharply by nearly 6%, with foreign investors net purchasing 1.7 trillion won in equities.

South Korea’s stock market closed nearly 6% higher on Thursday, rebounding from a sharp decline in the previous session, as chip stocks surged on the back of shareholder return measures and falling U.S. Treasury yields. Chipmaker SK Hynix rose 12.73% after announcing it would repurchase and cancel its treasury shares; peer Samsung Electronics gained 9.49%. Earlier media reports said Samsung Electronics will also unveil a new shareholder return policy by the end of this month. A total of 905 stocks changed hands on the day, with 521 advancing and 338 declining. Foreign investors net bought 1.7 trillion won (about $12.2 billion) worth of South Korean stocks. (Jinshi)

19 minutes ago

A whale has resumed live trading on Binance, setting 10 major goals first, and is currently shorting $222 million worth of BTC and ETH.

On-chain analyst Ai Yi (@ai_9684xtpa) tracked that the trader codenamed "Set 10 Big Goals" resumed live trading on Binance on July 27, after a one-month break. As of now, his total short positions in BTC and ETH contracts are valued at around $222 million, with an accumulated unrealized profit of roughly $401,000. For BTC, he holds 4x-leveraged short positions totaling 2,236.384 BTC, with a position value of approximately $156 million, an entry price of $69,826.87, and current unrealized profit of about $369,000. For ETH, he holds 6x-leveraged short positions totaling 29,316.677 ETH, with a position value of around $66.1 million, an entry price of $2,254.74, and current unrealized profit of roughly $32,000.

19 minutes ago

NeoSoul completes $11 million in Pre-A round financing, accelerating its expansion into the AI economy.

NeoSoul announces completion of $11 million Pre-A round financing. Participating institutions in this round include MH Ventures, Amber Group, ArkStream Capital, 0G Foundation, Kirin Capital, CatcherVC, and New Oak International. The capital will be used to further advance NeoSoul’s development of AI agent trading products and AI economic infrastructure. This financing round comes shortly after the launch of NeoTrade, NeoSoul’s AI agent trading workbench, which allows traders to configure their own AI trading agents, enabling these agents to make independent decisions and execute trades autonomously. AI is evolving from auxiliary analysis to independent execution; for trading, the industry’s focus is shifting to how to maintain agent autonomy while ensuring controlled fund security. This round brings together multiple investment institutions from digital assets, Web3 infrastructure, decentralized AI, Asian capital, and North American markets. Notably, Kirin Capital, a key investor in this round, has long been rooted in the Vietnamese and Southeast Asian capital markets, and its investment will further support NeoSoul’s expansion into Vietnam and Southeast Asia. NeoSoul co-founder Kaelan stated: “AI is moving from information production to autonomous economic activities, and trading is one of the earliest scenarios that can form a complete economic closed loop. NeoTrade serves as the entry point; following this financing round, NeoSoul will continue building infrastructure that supports large-scale participation of AI agents in economic activities.” NeoSoul plans to allocate the capital from this round to further develop NeoTrade, strengthen its trading infrastructure, and expand its global ecosystem. The company will continue advancing product development around the connection between autonomous AI decision-making and controlled fund execution.

19 minutes ago

Crypto options with a notional value exceeding $1.8 billion are set to expire and settle tomorrow, with BTC’s key pain point at $66,000.

According to Deribit data, crypto options with a total notional value of $1.82 billion are set to expire and settle tomorrow. Breakdown: BTC options have a notional value of $1.57 billion, with a put/call ratio of 0.65 and a max pain point of $66,000; ETH options carry a notional value of $250 million, a put/call ratio of 0.77, and a max pain point of $1,950.

19 minutes ago

The third-largest HYPE long position on Hyperliquid has an unrealized profit of $14.58 million, and has not taken any additional actions since opening a 10x long at a low level.

According to monitoring by TradingBeats (formerly Hyperinsight), address 0x6666...23f5 currently holds 450,000 HYPE long positions on Hyperliquid, with a position value of approximately $32.49 million. Its average entry price is $39.8114, and the current unrealized profit is around $14.58 million. On-chain fund flow data shows that the address’s initial funds entered HyperCore primarily via Circle CCTP-related HyperEVM CoreDepositWallet. The first deposit was made on March 18, 2026, followed by cumulative receipts of about 3.8 million USDC between March 18 and 27, all from the HyperEVM CoreDepositWallet. Subsequently, starting at 02:31 HK time on March 18, 2026, the address built a 10x leveraged HYPE long position from scratch, completed its last position addition on April 7, and has since been inactive (making it impossible to calculate its historical trade win rate). As of press time, the address has paid $950,000 in funding fees for its long positions, ranking third among all HYPE contract long addresses on Hyperliquid by position size.

19 minutes ago

Bitcoin surges past $70,000, with an 8.6% gain over the past 24 hours.

According to HTX market data, Bitcoin has broken through $70,000, posting an 8.6% gain over the past 24 hours.

19 minutes ago

Disney Stock Extends Rally as New Avengers Doomsday Trailer Drops

Citigroup: Gold and silver still have upside potential; improving geopolitical and macroeconomic conditions will revive demand for precious metals investments.

Citigroup Research stated that the uptrend in precious metals is far from over, with silver continuing to track gold’s moves and serving as a more aggressive play on upside due to its higher volatility. Should tensions in the Strait of Hormuz eventually ease, paired with a less hawkish stance from the Federal Reserve, investment demand for precious metals will continue to recover. Recent market action has already laid the groundwork for this outlook. The settlement price of COMEX August gold futures rose 0.49% to $4,383 per ounce; silver futures fell 0.5% on the day to $64.769 per ounce, ending a two-day winning streak. Citigroup noted that the short-term pullback does not alter silver’s status as a high-beta asset relative to gold, adding that if easing geopolitical risks drive capital back into precious metals, silver could rise to $95 per ounce by 2027. However, Citigroup also outlined risk scenarios, estimating there is roughly a 20% chance silver could drop to $50 per ounce, highlighting that current precious metals trading remains highly dependent on interest rate expectations, U.S. dollar movements, and geopolitical risks. For markets, gold remains a core asset for hedging and rate-cut expectations, while silver is better suited to capturing upside from a rebound in risk appetite.

18 minutes ago

South Korea plans to raise the trading threshold for single-stock leveraged ETFs in September, lifting the minimum trading unit to 20 shares.

The Korea Exchange (KRX) has drafted amendments to the Detailed Rules for the Implementation of the Business Regulations for the Securities Market, raising the minimum trading unit for single-stock leveraged ETFs from the current 1 share to 20 shares, and for ETNs from 1 security to 20 securities. The proposal is now open for public consultation. It is reported that South Korea’s financial regulators originally planned to roll out the measure in November, but given the rising speculative trading in single-stock leveraged products, a pre-emptive cooling is needed, so the exchange is pushing for the earliest possible implementation next month. In addition to raising the trading unit, South Korean regulators will also strengthen investor protection measures: new individual investors must complete at least 5 trading days of simulation trading (1 hour per day, totaling 5 hours) before participating in single-stock leveraged product transactions.

18 minutes ago

MetaPlanet transfers 1,473 $BTC ($93.82M), Hut 8 moves 493 $BTC ($31.36M)

MetaPlanet(@Metaplanet) transferred out 1,473 $BTC($93.82M) 1 hour ago. Hut 8(@Hut8Corp) also transferred out 493 $BTC($31.36M) 3 hours ago.

18 minutes ago

Bernstein raises Microsoft's price target to $660, with AI and cloud businesses set to continue underpinning earnings growth.

In its latest research report, Bernstein reaffirmed its bullish outlook on Microsoft, raising the 12-month target price to around $660 and maintaining an "Outperform" rating. The bank argues that market concerns over Microsoft’s AI infrastructure being "built too fast and spent too much" are overblown, noting that the company’s data center expansion remains aligned with its commercial cloud revenue growth, and AI investments will continue to support Azure, Copilot, and its enterprise software ecosystem. Microsoft’s share price has rebounded notably recently, rising roughly 30% since the release of strong earnings at the end of July, returning to positive territory for the year. Even after the rally, Microsoft’s valuation remains below its historical high range, with its AI and cloud businesses continuing to underpin earnings growth, per Bernstein. The report specifically points out that Microsoft’s long-term lease obligations total $329.1 billion, spread across fiscal years 2027 to 2033, with some contracts extending up to 20 years; hardware commitments for fiscal 2027 are approximately $169 billion, after which they will decline significantly. This means Microsoft is not locked into a single short-cycle AI hardware expenditure, and can reallocate data center resources to traditional cloud and software services in the future.

18 minutes ago

Korea Exchange activates KOSPI sidecar mechanism, suspends program buying orders.

Korea Exchange triggered the KOSPI temporary trading halt mechanism, suspending programmatic buying for 5 minutes, following a 5% rise in KOSPI 200 index futures. (Jinshi)

18 minutes ago

A $100 million BTC short position has emerged on Hyperliquid, with whale DoshiAtoll opening a 40x leveraged short.

According to monitoring by ai_9684xtpa, Hyperliquid trader DoshiAtoll opened a short position of 1,576.23 BTC at 8 AM today, valued at roughly $100 million. The short entry price was $64,039, with 40x leverage, and it currently has an unrealized profit of approximately $456,000. Following the establishment of this position, the two largest BTC holdings on the Hyperliquid platform are now both BTC short positions, drawing market attention to the bearish sentiment of large capital.

18 minutes ago

, /PRNewswire/ -- AITECH Cloud Network, formerly Solidus AI Tech, has announced a strategic partnership with Secret Network to integrate confidential computing capabilities into Agent Forge, its no-code AI agent platform developed fully by ACN's internal development team and launched in April 2025.

The collaboration will enable users of Agent Forge to build and deploy AI agents and workflows with enhanced privacy, verifiable execution, and confidential processing powered by Secret Network's confidential computing infrastructure.

The partnership represents a significant step towards addressing one of the biggest challenges facing enterprise AI adoption: how to utilise powerful AI systems while maintaining data privacy, security, and trust.

Under the agreement, Secret Network's confidential AI infrastructure will be integrated into Agent Forge's growing ecosystem, allowing developers and businesses to access secure AI models and confidential workflow execution directly from the platform. Initial integration efforts will focus on Agent Forge's light-mode chat interface, with plans to expand confidential execution capabilities across the broader workflow builder environment.

Agent Forge has rapidly expanded since private beta launch, offering users a no-code environment for building AI agents, workflow automations, and integrations through a growing library of templates and external services. The platform recently completed its migration to Ethereum, positioning it for broader institutional adoption.

The collaboration will also introduce verification tools that allow users to confirm that AI workloads are running within trusted confidential computing environments. By combining Agent Forge's agent orchestration capabilities with Secret Network's confidential infrastructure, organisations will be able to deploy AI workflows with greater confidence in data security and operational integrity.

John Mendez, Head of AI Development of AITECH Cloud Network, said:

"Agent Forge was built to make AI agent creation accessible to everyone, from first-time users to enterprise teams. As AI adoption accelerates, privacy and trust become critical requirements. Our partnership with Secret Network allows us to introduce confidential AI capabilities directly into the platform, giving users the ability to build, deploy, and verify secure AI workflows without adding complexity to the user experience."

Luke B, Chief of Operations at Secret Network Foundation, said:

"Confidential computing is becoming a foundational layer for the next generation of AI applications. By integrating Secret Network's confidential infrastructure into Agent Forge, we are enabling developers and businesses to leverage advanced AI while maintaining control over sensitive data. Together, we are helping establish a future where AI systems are both powerful and verifiable."

The companies expect the initial integration to focus on confidential model access and verification features, followed by the deployment of confidential execution environments capable of supporting fully private AI workflows. Future collaboration will include joint community initiatives, educational content, and ecosystem development activities.

The partnership reflects a broader industry shift towards confidential AI, where privacy-preserving technologies and trusted execution environments are increasingly viewed as essential components of enterprise-grade AI infrastructure.

About AITECH Cloud Network

AITECH Cloud Network, formerly Solidus AI Tech, is a provider of AI and high-performance computing infrastructure. Its ecosystem includes Agent Forge, a no-code AI agent platform developed by ACN'steam, a competitor to n8n and make.com, that enables users to build, deploy, and manage intelligent agents and workflow automations, through a conversational interface.

About Secret Network

Secret Network is a leading confidential computing platform that enables private and secure applications through trusted execution environments and privacy-preserving infrastructure. The network provides developers with tools to build confidential AI, secure data applications, and verifiable computing solutions.

SOURCE Secret Network Foundation

, /PRNewswire/ -- Morpheus, a decentralised marketplace for AI inference, has announced a partnership with Secret Network to enable private, secure execution of AI models using Trusted Execution Environments (TEEs).

The collaboration introduces a new TEE Provider developed by Morpheus, capable of running confidential AI models within Trusted Execution Environments. Through this integration, Secret Network's Confidential AI models will be accessible to Morpheus users, allowing AI inference to be performed without exposing sensitive data or model interactions.

AI inference, the stage at which trained models process new inputs to generate outputs, represents the primary point of interaction between users and AI systems. It is also where data privacy risks are most acute, particularly in decentralised environments where compute is distributed across multiple nodes. Any industry edge? No longer secret. No longer yours.

By combining Morpheus's decentralised infrastructure with Secret Network's privacy-preserving technology, the partnership aims to address this challenge directly.

"This is about making decentralised AI usable at an enterprise level," said David Johnston of Morpheus. "Inference is where real-world value happens, but it is also where trust breaks down. By enabling confidential execution through TEEs, we are ensuring that users no longer have to choose between openness and privacy."

Secret Network's technology ensures that data remains encrypted not only in storage and transit, but also during computation. Within a TEE, inputs are processed securely, with neither the node operator nor external parties able to access the underlying data or outputs.

"Privacy cannot be an afterthought in AI," said Luke B, COO at Secret Network Foundation. "As inference becomes the dominant mode of interaction with AI systems, protecting that layer is essential. This partnership brings confidential AI into a decentralised marketplace, which is a significant step forward for both adoption and trust."

The integration reflects a broader shift towards confidential computing in AI, where sensitive data can be processed securely even in distributed environments. It also signals growing demand for infrastructure that supports both scalability and privacy, particularly as AI applications expand across finance, healthcare and enterprise systems.

Through this partnership, developers and enterprises using Morpheus will be able to access Secret Network's Confidential AI models with full privacy guarantees, enabling new use cases that require secure, decentralised intelligence.

Stop theft. Stop worrying.

About Morpheus
Morpheus is a decentralised marketplace for AI inference, connecting users and developers with distributed compute resources to run AI models at scale.

About Secret Network
Secret Network is a blockchain platform focused on privacy-preserving computation, enabling confidential smart contracts and secure data processing through Trusted Execution Environments.

SOURCE Secret Network Foundation

A federal prosecutor indicted an FBI agent for the alleged theft of almost $1 million worth of cryptocurrency. The money reportedly originated from government-controlled crypto wallets. Investigators reported that he used ChatGPT to plan investments and relocate to Europe. The Department of Justice has indicted a supervisory special agent of the FBI for theft of nearly $1 million in cryptocurrency during the investigation of a crime. According to court filings, Patrick Steven Yaroch, a counterintelligence agent with a Top Secret clearance, allegedly abused his official position to move cryptocurrencies stored in the wallets being investigated by the government. In addition, it is alleged that he later used the assistance of ChatGPT to investigate investment opportunities, residency possibilities in Europe, and financial planning after having moved the money. As pointed out by the authorities, the charges against Mr. Yaroch are still pending; however, he enjoys the benefit of innocence at this point.

Alleged Thief Gained Access Internally  According to the federal affidavit presented in the Eastern District of Virginia, Patrick Steven Yaroch worked in the FBI’s Counterintelligence and Espionage Division after joining the bureau in 2017. The prosecutors believe that he had access to cryptocurrency wallets associated with the adversarial nation. He learned recovery seed phrases, set up his own wallets, and moved assets in various transfers in 2024 and 2025. 

The investigators have managed to recover about $925,426 worth of the funds with ongoing efforts to trace more money. Yaroch voluntarily contacted Department of Justice officials and admitted making poor decisions involving cryptocurrency. Investigators reported that he revoked consent to recover the wallet seed phrases and requested legal counsel.

Searches for ChatGPT Conversations Extend Investigation The investigators managed to extract the ChatGPT conversations from Yaroch’s mobile in the course of their investigation. Yaroch asked questions regarding his plans to invest $1 million and move to Europe. With his discussions involving Portugal, Italy, Turkey, and Greece. In addition, the investigators uncovered travel reservations for Portugal. Along with documents empowering him to act on behalf of the Portuguese tax authorities. It is also alleged that Yaroch transferred approximately $1.02 million to Suilend, a lending protocol on the Sui blockchain. While investigators found more cryptocurrency stored in his Kraken wallet.

Highlighted Crypto News:
Mastercard Completes $1.8B BVNK Acquisition to Expand Stablecoin Payments

I specialize in Web3 and crypto writing, producing clear, research-driven content on blockchain, cryptocurrencies, and market trends.

In brief OpenAI published a detailed rebuttal to Apple's trade secrets lawsuit. The company says Apple misstated key facts before filing suit and released emails to support its claims. OpenAI also argues Apple employees continued seeking help from a former engineer after he joined the company. OpenAI has publicly challenged Apple's trade secrets lawsuit, publishing emails, text messages, and a point-by-point rebuttal that accuses the iPhone maker of making factual errors before filing its complaint.

In a blog post published Monday, OpenAI said Apple confused two employees with similar names, falsely claimed the company ignored its outreach, and incorrectly asserted that Apple's outside counsel had spoken with OpenAI General Counsel Che Chang before filing a lawsuit against OpenAI.

The email and text dump comes after Apple sued OpenAI in July, accusing the company, former Apple engineer Chang Liu, former Apple design executive Tang Tan, and io Products of misappropriating hardware trade secrets as OpenAI expanded its consumer hardware efforts following its acquisition of Jony Ive's startup in 2025.

“Apple is one of the greatest companies of all time, and built a reputation for obsessing over the smallest details,” OpenAI wrote. “This careless, aggressive and oddly personal lawsuit sadly doesn’t live up to that reputation.”

In its lawsuit, Apple accuses former Apple engineer Chang Liu of retaining a company laptop after leaving, exploiting an authentication bug to access Apple's internal systems, and downloading confidential engineering documents. The lawsuit also claims former Apple executive Tang Tan used confidential Apple information to benefit OpenAI's hardware efforts and that the company's recruiting process encouraged prospective hires to bring sensitive Apple materials.

OpenAI disputes those allegations, saying Apple contacted the wrong person after confusing two employees with similar surnames and incorrectly claimed its lawyers had spoken with OpenAI General Counsel Che Chang before filing suit.

To support its case, OpenAI published iMessage exchanges between Liu and Apple employees sent after he left the company in January, showing them asking for help locating files, transferring documents, and answering technical questions—conversations OpenAI says undermine Apple's claims that Liu improperly retained access to confidential information.

According to OpenAI, those exchanges show Apple employees relied on Liu's access and that any remaining access resulted from Apple's own offboarding process rather than deliberate misconduct.

As for Tan, OpenAI says the former Apple executive was repeatedly instructed not to bring or use confidential information from his previous employer. The company also argues Apple's request for a preliminary injunction, which could restrict OpenAI's activities while the case proceeds, is unnecessary.

“Apple’s request for a preliminary injunction is both based on false information and completely unnecessary because we do not have, nor want, any of their trade secrets,” the company wrote. “We’re much more interested in building innovative products and technologies that push the frontier.”

Daily Debrief NewsletterStart every day with the top news stories right now, plus original features, a podcast, videos and more.

Table of contents

The arrest of an FBI intelligence agent with top-secret clearance on charges of stealing cryptocurrency from wallets the bureau had been investigating raises an uncomfortable question: who guards the guardians when the asset is a bearer instrument with no reversibility?

According to the original report, the agent worked inside FBI headquarters and had access to digital wallets that were part of active criminal investigations. The theft, totaling roughly $1 million, was discovered during an internal review, leading to the agent’s arrest. The case highlights a growing vulnerability for law enforcement agencies that increasingly seize and hold cryptocurrency — often for months or years as cases grind through the courts.

A Breach from Inside the Bureau The agent’s top-secret clearance and placement at headquarters are the most jarring details. These positions come with layers of vetting, background checks, and ongoing monitoring. That an individual with that level of access could allegedly siphon funds from wallets the FBI itself had targeted suggests the controls around private key management were not airtight.

Seized crypto typically moves to a government-controlled wallet under multi-signature arrangements, though exact procedures vary by agency. The FBI has not disclosed how the theft occurred, whether it involved unauthorized transactions, compromised seed phrases, or exploitation of procedural gaps. What is clear is that the funds could be traced on-chain, which may have accelerated the internal investigation. Unlike cash, crypto ledger activity is permanent and public. That traceability likely left a digital trail even a trained intelligence agent could not completely erase.

Custody Risks for Seized Crypto The incident forces a hard look at how federal agencies handle digital assets under seizure. Unlike traditional assets that sit in evidence lockers or frozen bank accounts, cryptocurrency requires active key management, access controls, and technical safeguards that many government units are still learning to implement. The standard for most private custodians — hardware security modules, geographic distribution of keys, mandatory multi-sig execution — is not always replicated inside bureaucratic structures built for physical evidence.

Every seized wallet sitting on a government ledger is a temptation point for an insider who understands that moving coins to a mixer or an exchange can erase the trail in minutes. The FBI has recovered stolen funds in past cases only because the thief made operational mistakes. This time, that appears to have happened again. But the outcome could easily have been worse.

A Pattern of Insider Theft This is not the first time a U.S. law enforcement official has been accused of stealing crypto from evidence. In 2021, a Secret Service agent was sentenced for stealing Bitcoin during the Silk Road investigation. DEA agents have faced similar charges in prior years. Each case revealed the same structural weakness: the people trusted to investigate crypto crime are often the same people capable of committing it, using the same tools and knowledge they deploy against suspects.

The news arrives as lawmakers in Washington debate a sweeping crypto bill, with traditional financial heavyweights attempting to water down legislation, according to recent coverage. Incidents like the FBI arrest give critics of government crypto oversight fresh arguments: if the FBI cannot secure its own investigative wallets, what confidence should the public have in broader federal digital asset mandates?

Broader Implications for Crypto Enforcement The arrest will likely trigger a round of internal audits across multiple agencies that hold seized crypto. The Department of Justice may push for standardized key custody protocols, possibly requiring third-party qualified custodians or technology partners that already serve the private sector. That would be a significant shift from current practice, where agencies often self-custody.

For defendants in crypto-related criminal cases, the theft also creates an avenue to challenge evidence integrity. If the FBI cannot demonstrate an unbroken chain of custody for digital assets, defense attorneys may argue that funds attributed to their clients could have been manipulated or replaced. The legal precedent is thin, but the concept is straightforward. A single compromised wallet can cast doubt across an entire investigation.

The case also underscores the human perimeter problem. No amount of encryption or multi-sig eliminates the risk that an authorized signer goes rogue. The remedy lies in separation of duties, real-time monitoring of wallet activity, and penalties severe enough to deter the next insider. Right now, that deterrent is failing.

AUTHOR

Farhan Karim is a technology writer and content strategist with 15+ years of experience writing thousands of articles, blogs, whitepapers, and ebooks on Blockchain, Cryptocurrency, and other tech niches. His expertise in content strategy, SEO, and a keen eye on the ever-evolving tech space have led him to work with companies like Pepsi, Huawei, Arab News, and now Blockchain Reporter.

US tech giants have cut nearly 140,000 jobs this year, with the four leading players' AI capital expenditure totaling $725 billion.

According to statistics from the Financial Times in partnership with Challenger, Gray & Christmas, U.S. tech industry layoffs since 2026 have accounted for more than one-third of all announced layoffs nationwide. Amazon, Oracle, Meta and Microsoft alone have cut nearly 50,000 jobs, roughly 6% of their total workforce. In sharp contrast, Amazon, Alphabet, Meta and Microsoft are projected to invest a combined $725 billion in AI infrastructure such as data centers this year. After laying off staff in March, Oracle’s total headcount dropped by 21,000 for the full year; this month, S&P downgraded its credit rating, citing weak cash flow and uncertain AI returns. Microsoft cut 4,800 jobs this month, mainly in its Xbox gaming division, essentially a full reset of its $75 billion acquisition of Activision Blizzard three years ago. The narrative that "AI causes layoffs" is met with skepticism in academic circles. Enrico Moretti, an economics professor at the University of California, Berkeley, notes that AI-related layoffs are more of an excuse for management to correct over-hiring during the pandemic. "Claiming AI-driven efficiency gains is easier than admitting to over-hiring back then," he said. Market pricing also contradicts this narrative: within 30 trading days of announcing layoffs, companies that attributed cuts to AI saw their stock prices underperform the Nasdaq by nearly 10%, while companies laying off for other reasons lagged by only around 4%. Amazon and Microsoft have explicitly stated that AI adoption is not a decisive factor in their layoffs. In contrast to the tech giants’ non-core business contractions, AI-native startups like Anthropic and OpenAI are still rapidly expanding their workforce, driving fast growth in AI sector employment. "What is being cut is merely all other non-core business segments."

27 minutes ago

Layer1 project Vanar will begin migrating its infrastructure to Base next Tuesday.

Layer 1 blockchain project Vanar announced that its infrastructure migration to Base will kick off next Tuesday. Users currently staking VANRY must first unstake, wait for the cooldown period to elapse before claiming their tokens. Earlier, Vanar stated that existing VANRY token holders can complete the migration at a 1:1 ratio, with their holding amounts remaining unchanged. Additionally, VANRY’s total supply will rise from 2.4 billion to 10 billion tokens, approximately 62% of which will stay locked during the migration. Once the migration is complete, staking for Vanarchain validators will be halted.

27 minutes ago

2035年数据中心将占美国电力消耗的约20%，成为下一个AI瓶颈

U.S. data center power demand is projected to surge by 253% from 2026 levels, reaching a record 194 gigawatts by 2035 — with 1 gigawatt roughly matching the capacity of a traditional nuclear reactor. Currently, data centers consume 6% of the U.S.’s annual electricity; that share is estimated to climb to around 12% by 2030, and will account for roughly 20% of total U.S. electricity consumption by 2035. Most of the growth in U.S. power demand is concentrated in a handful of grid regions, such as the PJM Interconnection, which serves Washington, D.C. and 13 states including Virginia, Pennsylvania and Ohio. Power will be the next AI bottleneck.

27 minutes ago

Robinhood Chain's 24-hour network fee revenue reached $350,000, ranking fourth among all blockchains.

According to DeFiLlama data, Robinhood EVM Chain generated $350,000 in 24-hour network fee revenue, ranking fourth among all blockchains, trailing only Canton, Tron, and Solana. Launched on July 1, the Robinhood EVM Chain has seen its total value locked (TVL) quickly rise to $315 million. Though originally designed for on-chain stocks and ETFs, it has emerged as a major hub for meme coin activity.

27 minutes ago

Citrini’s view: Korean manufacturers hold a cautious stance on the commercialization of 3D IC, while China’s ChangXin Memory Technologies is accelerating to secure niche markets via a customized approach.

Citrini analyst Jukan, citing the latest industry analysis from ZDNet Korea, notes that the AI semiconductor performance race is shifting from process miniaturization to advanced stacking technology, with 3D ICs—vertically integrating logic and memory—emerging as a key next-generation technology. Samsung Foundry has recently seen a sharp surge in inquiries for 3D ICs, with nearly every potential client asking about the technology. However, Jukan argues that 3D ICs are essentially custom DRAMs built to order, which fundamentally conflicts with the core business models of Samsung and SK Hynix, both relying on mass-producing standardized products. The two Korean giants are expected to limit their involvement to preliminary research and remain cautious about full commercialization. This structural hesitation opens an opportunity for niche players. China’s ChangXin Memory Technologies (CXMT), barred from entering the HBM market due to U.S. semiconductor restrictions, is leveraging 3D ICs as a differentiating breakthrough, choosing to target custom memory instead of competing head-on with Korean firms in the bulk DRAM segment. A South Korean semiconductor industry official stated that China has taken the lead in producing multiple 3D DRAM sample chips, and niche market players like CXMT and Winbond are likely to be the first to build this market. Analyst Jukan has also previously emphasized that NVIDIA’s CUDA moat is ending: the erosion of its software ecosystem barrier is occurring in tandem with the trend of 3D IC customization, and the next-generation AI memory competitive landscape may face a structural reshuffle.

27 minutes ago

A whale holding a 40x leveraged long Bitcoin position worth nearly $40 million was stopped out after just 18 hours.

According to on-chain analyst Ai Yi (@ai_9684xtpa), a whale that opened a 40x leveraged long BTC position worth $38.67 million starting last night cut its losses after holding the position for just 18 hours, suffering a single loss of $368,000. The whale’s entry price was $64,469, with a stop-loss set at $63,859, triggering an immediate stop-loss after BTC dropped by just 1%.

27 minutes ago

WSJ: Treasure Global's Subsidiary Tadaa Capital Acquires Cigar Secret to Develop an AI-Power Premium Lifestyle Membership Ecosystem