New models and updates from major AI providers this week
Cohere has released Cohere Transcribe Arabic, an open-source speech-to-text model fine-tuned for high accuracy in Arabic dialects. The 2B parameter conformer-based model is optimized for production throughput and supports both Arabic and English accents.
IBM researchers have introduced CoFrGeNets, a new architecture designed to replace the traditional 'bones' of transformer-based models. This research focuses on structural improvements to generative AI architectures.
NVIDIA has released the Nemotron 3 Ultra, which achieves benchmark-leading performance and business task parity with top closed models. By tuning the LangChain Deep Agents harness around the model rather than retraining it, NVIDIA provides a high-throughput solution at 1/10th the inference cost of leading competitors.
xAI has introduced Grok 4.5, positioned as the company's smartest model to date. It is specifically optimized for coding tasks, agentic workflows, and complex knowledge work.
StepFun has released Step 3.7 Flash, a 198B-parameter sparse Mixture-of-Experts (MoE) vision-language model designed for high-frequency production workloads and agentic workflows. The model features native image understanding via a 1.8B visual encoder and offers selectable reasoning levels to balance speed, cost, and depth.
Anthropic has transitioned Claude Code from an internal command-line interface to a fully realized coding agent. The release provides insight into its development as a specialized tool for software engineering tasks.
OpenAI has introduced GPT-Live, a new product designed for real-time interaction. The release includes accompanying safety documentation via the GPT-Live System Card to address deployment considerations.
xAI has launched new flagship Grok Voices, introducing a variety of multilingual options for the Grok Voice interface. The update also focuses on enhancing the naturalness and conversational quality of the five original voices.
Selected arXiv and HuggingFace papers this week
Paper 1
The paper introduces Direct On-Policy Distillation (Direct-OPD) as a method to transfer reasoning capabilities from large models to smaller models by focusing on policy shifts rather than absolute policies.
TL;DR
This technical paper proposes a new distillation paradigm called Direct-OPD to improve weak-to-strong generalization in reasoning models. By distilling only the policy shift induced by reinforcement learning, the method avoids the capacity limitations inherent in traditional teacher-student imitation.
The research addresses a critical bottleneck in the development of frontier reasoning models: how to effectively transfer high-level reasoning capabilities from large 'teacher' models to smaller, more efficient 'student' models. Current methodologies rely heavily on On-Policy Distillation (OPD), where students are trained on states sampled under teacher supervision. However, existing OPD methods and their hybrids often suffer from a capacity ceiling; if the teacher model is used to imitate an absolute policy, the student effectively inherits the limitations of that teacher. Furthermore, research indicates that small models struggle significantly when attempting to copy much stronger reasoners directly. To solve this, the paper introduces Direct On-Policy Distillation (Direct-OPD). Unlike standard approaches that attempt to mimic the teacher's final output distribution, Direct-OPD focuses exclusively on the policy shift—the difference between the post-RL policy and the reference policy (log πT - log πTref). By discarding the absolute policy and focusing only on the improvements introduced by reinforcement learning, the student model can more effectively learn the underlying reasoning logic without being constrained by the teacher's absolute performance bounds. This technique facilitates better weak-to-strong generalization, allowing smaller models to potentially extrapolate beyond the immediate imitation of the teacher's final state.
Paper 2
The article introduces KVpop, a novel method for efficient Key-Value (KV) cache compression in large language models using predictive online pruning.
TL;DR
This technical paper presents KVpop, a technique for compressing the KV cache in transformer models via predictive online pruning. It optimizes the computation of attention targets by reusing sparse log-normalizers and utilizing efficient data structures like Fenwick trees.
The research introduces KVpop, an advanced approach to Key-Value (KV) cache compression designed to maintain model performance while significantly reducing memory overhead during inference. The core innovation lies in its ability to predictively prune the KV cache using a student-teacher framework. To avoid the massive computational cost of calculating dense causal attention probabilities, KVpop implements a transposed-attention target computation. This method swaps query and key roles in an efficient attention kernel, such as FlexAttention, to approximate future-attention mass without materializing large S×S matrices. By reusing sparse log-normalizers from the student pass, the system achieves high accuracy with minimal added inference-time overhead. Furthermore, the algorithm manages a top-k sparse attention pattern by maintaining a union of sink tokens, a recent window, and high-priority tokens. To keep this process efficient as the sequence length grows, the authors utilize a Fenwick tree to track token ranks, allowing for a query-specific cutoff computation in O(S log S) time. This ensures that the sparse mask is generated dynamically within fused kernels, optimizing both space and time complexity.
Paper 3
The introduction of ReChannel, a method that repurposes text-to-image Diffusion Transformers (DiTs) for dense prediction by reading out task-native pixel-space fields directly instead of generating RGB targets via VAE decoders.
TL;DR
This paper proposes ReChannel, a novel architecture that transforms text-to-image models from RGB generators into efficient dense prediction engines. By treating transformer tokens as spatial carriers for task-specific data rather than RGB pixels, the method achieves new state-of-the-art performance with much higher computational efficiency.
The research addresses a fundamental inefficiency in using large-scale text-to-image models for dense prediction tasks like depth estimation, segmentation, and matting. Current approaches typically treat these tasks as an image-to-image translation problem, encoding task targets into an RGB-trained VAE latent space and decoding them back to pixels. The authors argue that this 'generation' interface is unnecessary because dense prediction requires pixel-accurate task fields, not the reconstruction of complex RGB textures. They propose 'ReChannel,' which leverages the inherent patch-based spatial structure of Diffusion Transformers (DiTs). In ReChannel, the pretrained DiT acts as a field organizer; the input passes through a standard VAE encoder, but the output bypasses the decoder entirely. Instead, a lightweight token-local linear head maps adapted tokens directly to task-native pixel patches. This approach was validated using the FLUX-Klein backbone across six different dense prediction tasks and over twelve benchmarks. The results show that ReChannel not only reaches state-of-the-art performance in areas like KITTI depth and trimap-free matting but also provides a massive speedup, performing up to 2.48x faster than previous generative editing methods while using significantly fewer parameters for the output head.
Paper 4
The research explores an analysis-driven approach to transformer linearization, specifically identifying why delta-style state updates better approximate softmax attention than gated accumulation.
TL;DR
This paper presents a method for converting pretrained transformers into linear-time architectures by focusing on the efficiency of state update designs. By analyzing softmax attention through a first-order approximation, the authors prove that delta-style updates are superior for post hoc linearization.
The research addresses the quadratic computational bottleneck of causal self-attention in large language models, which limits long-context inference. The authors propose an analysis-driven approach to 'post hoc linearization,' a process where existing full-attention models are converted into linear-time architectures without requiring retraining from scratch. To avoid the confusion caused by simultaneous interventions like LoRA or distillation, the study employs a controlled experiment with a frozen backbone, training only the new replacement mechanism parameters.
A key theoretical contribution is the derivation of a first-order approximation that connects softmax attention to linear state updates, revealing that softmax relies on key-dependent, rank-1 orthogonal projections. This finding explains why delta-style networks (such as Gated Delta Networks) outperform purely gated accumulation models (such as Gated Linear Attention), as the former can implement necessary geometric corrections. To further close the performance gap, the researchers introduce several structural interventions: sink tokens to handle attention spikes, short convolutions for projection adaptation, and a disjoint sliding-window path for local context preservation.
The effectiveness of this approach was validated by scaling it across LLaMA and Qwen models up to 32B parameters. The results show that this linearization method outperforms prior post hoc baselines on the MMLU benchmark and achieves long-context retrieval performance comparable to much more complex adaptive-caching frameworks.
Paper 5
The article introduces DSpark, a novel confidence-scheduled speculative decoding method that utilizes semi-autoregressive generation to accelerate large language model inference.
TL;DR
This technical paper presents DSpark, a new approach to speculative decoding designed to enhance the speed of autoregressive generation. It contextualizes the work within the broader landscape of drafting architectures and parallel generation strategies.
The article explores the evolution of speculative decoding, a technique used to accelerate the inference of large language models by separating the proposal of tokens from their verification against a target model. The efficiency of this process is highly dependent on the accuracy and speed of the 'drafter.' Historically, researchers have utilized small standalone language models as drafters, but recent innovations have integrated multi-token heads or feature extrapolators directly into the target model's architecture to improve performance. Other specialized strategies mentioned include self-speculation through early exits, dynamic vocabulary compression, prompt lookup, and retrieval-based methods. To address the sequential bottleneck inherent in traditional drafting, newer research has moved toward parallel or blockwise generation techniques such as Medusa, P-EAGLE, PARD, DART, and DFlash. Furthermore, some advanced methods like DDTree, TAPS, and JetSpec expand the draft chain into verifiable trees to maximize throughput. The paper positions DSpark within this landscape, highlighting its use of confidence-scheduled semi-autoregressive generation as a way to optimize the balance between drafting speed and verification accuracy.
Top stories curated from across the web this week
Article 1
A startup named Eigen Labs used AI agents and crowdsourcing to replicate and surpass Google's research on breaking elliptic curve cryptography using quantum computers.
TL;DR
Researchers at Google recently discovered a way to break 256-bit ECC encryption with much lower qubit counts than expected, but withheld their exact method using zero-knowledge proofs. The startup Eigen Labs successfully bypassed this secrecy by using AI agents and crowdsourcing to develop an even more efficient attack circuit.
The landscape of quantum computing security has been disrupted by recent developments in algorithmic efficiency. Google Quantum AI researchers recently published findings suggesting that 256-bit Elliptic Curve Cryptography (ECC)—a cornerstone of modern internet security and cryptocurrency—could be broken using only 1,200 to 1,450 logical qubits. To mitigate the risk of providing a roadmap for attackers, the Google team utilized a zero-knowledge proof to verify their results without disclosing the underlying mechanics. However, this strategy of information concealment was circumvented by the Seattle-based startup Eigen Labs. Using a combination of crowdsourcing and autonomous AI agents designed to analyze scientific literature and optimize quantum circuits, Eigen Labs' engineers were able to match Google's results within eight hours and surpass them within three days. Their final discovered circuit is approximately 47.5% more efficient than the one proposed by Google. This event highlights a growing tension in cybersecurity research between responsible disclosure and the rapid advancement of automated cryptanalysis tools. Furthermore, the significant reduction in required physical qubits serves as a stark warning for governments and organizations currently tasked with transitioning to post-quantum cryptography (PQC) standards before 2030.
Article 2
Discovery of the 'GitLost' vulnerability, an indirect prompt injection attack in GitHub Agentic Workflows that allows unauthorized access to private repositories.
TL;DR
Noma Labs has uncovered 'GitLost', a vulnerability in GitHub Agentic Workflows that enables indirect prompt injection. By posting malicious instructions in a public issue, an attacker can trick the AI agent into leaking sensitive data from private organizational repositories.
The research conducted by Noma Labs reveals a significant security flaw dubbed 'GitLost' within GitHub’s recently launched Agentic Workflows. These workflows leverage AI agents powered by models like Claude and GitHub Copilot to automate repository tasks using natural language. The core of the vulnerability lies in an indirect prompt injection attack: because the agent processes user-generated content—such as Issue titles and bodies—as part of its instruction context, it cannot distinguish between legitimate developer commands and malicious instructions hidden in text. An unauthenticated attacker can create a public GitHub Issue containing specifically crafted English instructions that command the agent to fetch files from private repositories and post them back to the public issue via comments. The researchers demonstrated that even when GitHub implemented guardrails to prevent such leaks, they were able to bypass these protections using simple linguistic manipulation, such as adding the word 'Additionally' to reframe the model's output. This discovery highlights a systemic risk in agentic AI systems where the context window serves as an expanded attack surface. To mitigate this, Noma Labs recommends that developers strictly scope agent permissions, never treat user-controlled content as trusted input, and implement rigorous sanitization of all inputs passed to LLM-based agents.
Article 3
Major technology vendors Microsoft, Google, and Cloudflare have accelerated their deadlines for implementing quantum-safe cryptography to 2029 due to the rising threat of quantum computing.
TL;DR
Tech giants are moving up their quantum-safe security deadlines to 2029 to combat the accelerating threat of quantum decryption. Organizations are urged to begin cryptographic inventories and implement crypto-agility immediately to protect against 'harvest now, decrypt later' attacks.
The landscape of cybersecurity is facing a significant shift as major technology players including Microsoft, Google, and Cloudflare have officially moved their target date for implementing quantum-safe cryptography forward to 2029. This change follows previous government directives from the US and France that aimed for a 2030 deadline. The primary driver behind this acceleration is the increasing realization that cryptographically relevant quantum computers may emerge much sooner than expected, necessitating a more urgent response to protect critical infrastructure.
Microsoft Azure CTO Mark Russinovich emphasizes that preparing for this transition is a massive, multi-year engineering challenge. Microsoft is integrating post-quantum cryptography (PQC) requirements into its Secure Future Initiative (SFI). Meanwhile, Google is already working on integration, such as implementing ML-DSA digital signature protection in Android 17. While current quantum technology remains in the Noisy Intermediate-Scale Quantum (NISQ) era—exemplified by chips like IBM Heron and Google Willow—the progress in various computing architectures, including neutral atoms, keeps the threat of 'Q-Day' a pressing reality.
A critical component of this risk is the 'harvest now, decrypt later' phenomenon. Security experts, such as Simon Pamplin from Certes, warn that state-level actors are currently intercepting and storing encrypted data with the intention of decrypting it once sufficiently powerful quantum computers become available. This means that even if a quantum computer does not exist today, the data being transmitted now is already at risk.
To mitigate these risks, experts recommend that organizations move beyond simply updating infrastructure to adopting a data-centric approach. The focus should be on creating 'living cryptographic inventories' to identify where encryption exists across applications and legacy systems. Ultimately, the goal for software architects and engineers is to build 'crypto-agility' into new products, ensuring that future transitions to new mathematical standards—such as lattice-based or hash-based structures—can be handled as routine updates rather than emergency security crises.
Article 4
The technical implications and mechanisms of the proposed SIGHASH_ANYPREVOUT flag for Bitcoin transactions.
TL;DR
This article explores the technical mechanics of the SIGHASH_ANYPREVOUT proposal and its ability to allow signature reuse across compatible UTXOs. It details how removing outpoint commitment facilitates Layer-2 protocols while highlighting potential risks like signature replay.
The article provides a deep dive into the proposed BIP 118, which introduces SIGHASH_ANYPREVOUT as a soft-fork upgrade for Bitcoin. Unlike existing SIGHASH modes like SIGHASH_ALL or SIGHASH_SINGLE, which cryptographically bind a signature to a specific outpoint (transaction ID and output index), ANYPREVOUT excludes the outpoint from the digest. This allows a single pre-signed transaction to be applied to any UTXO that meets the remaining criteria, such as matching amounts or scriptPubKey. The text distinguishes between two primary variants: ANYPREVOUT, which maintains commitment to the previous output's amount and script, and ANYPREVOUTANYSCRIPT, which removes these commitments entirely. While this flexibility is highly beneficial for Layer-2 scaling solutions and certain recovered-key constructions, it introduces the risk of signature replay attacks. If a signer does not carefully manage the remaining committed fields, an attacker or miner could potentially reuse a signature to spend a different UTXO than intended. Ultimately, while ANYPREVOUT enhances transaction expressiveness, it does not implement full recursive covenants on its own but serves as a foundational component for more complex script logic.
Article 5
A new three-stage Bluetooth attack tool named Whisper_Bully exploits CVE-2025-36911 to extract permanent addresses, execute DoS attacks, and hijack Fast Pair devices.
TL;DR
The Whisper_Bully tool demonstrates a sophisticated three-stage attack against Bluetooth Fast Pair devices. It leverages a specific vulnerability to bypass privacy randomization, followed by L2CAP flooding and connection hijacking.
Whisper_Bully is a technical security tool designed to demonstrate a multi-stage exploitation process against Bluetooth Low Energy (BLE) devices utilizing the Google Fast Pair protocol. The attack sequence begins with Stage 1, which targets CVE-2025-36911; by writing forged pairing requests and fake Account Keys to specific GATT characteristics (UUID 1236 and 1238), the attacker forces the device to transition from a randomized private address to its permanent factory-programmed BD_ADDR. Once the identity is known, Stage 2 initiates an L2CAP denial-of-service attack via 'EMP mode,' which involves sending rapid bursts of echo packets and intentional connection closures to exhaust the target's resources and lock out legitimate users. Finally, Stage 3 leverages the unstable state of the device during the flood to establish an unauthorized L2CAP connection, potentially allowing for reconnaissance or credential theft. The implementation is Python-based and relies on Linux-native Bluetooth stacks like BlueZ and tools such as bluetoothctl.
Article 6
Introduction of TorchJD, a PyTorch-based library designed for multi-task learning through Jacobian descent and scalarization techniques.
TL;DR
TorchJD is a specialized PyTorch library for optimizing neural networks with multiple simultaneous loss functions. It implements advanced methods like Jacobian descent and scalarization to resolve optimization conflicts in multi-task learning.
TorchJD is an open-source library built on top of PyTorch, specifically engineered to address the complexities of multi-objective optimization in neural networks. When training models with multiple tasks, standard gradient descent often struggles with conflicting gradients between different loss functions. TorchJD offers two primary strategies to mitigate this: scalarization and Jacobian descent. Scalarization methods, such as geometric mean or softmax weighting, merge multiple losses into a single scalar value before backpropagation. In contrast, the library's core strength lies in its implementation of Jacobian descent, which computes the Jacobian matrix of losses relative to parameters and uses advanced aggregators like UPGrad, MGDA, and CAGrad to determine an optimal, conflict-free update direction.
To ensure usability for large-scale deep learning, TorchJD introduces two specialized engines. The 'autojac' engine provides a familiar interface to developers accustomed to torch.autograd, allowing for the computation of Jacobians and the accumulation of these matrices in parameter fields. For memory-intensive applications, the 'autogram' engine allows for the incremental computation of the Jacobian's Gramian, bypassing the need to store full, massive Jacobian matrices in memory. This makes the library highly applicable to modern, large-scale architectures. The project is currently in beta and follows semantic versioning, providing a robust toolkit for researchers working on multi-task learning and multi-objective optimization.
Article 7
Analysis of various techniques used by threat actors to abuse Windows Services for persistence and privilege escalation.
TL;DR
This technical report details multiple methods for abusing Windows Services to achieve persistence and elevated privileges. It covers everything from traditional binary path modification to advanced techniques involving SDDL manipulation and service recovery hijacking.
The article provides a comprehensive breakdown of the 'Purple Team' approach to Windows Service exploitation. It identifies several primary attack vectors: service creation via native utilities like sc.exe and PowerShell, registry-based service injection, and binary path modification. A significant portion of the analysis is dedicated to more sophisticated methods, such as abusing the Security Descriptor Definition Language (SDDL) to alter permissions and exploiting the Windows Service recovery function to trigger malicious payloads upon service failure. The author highlights tools like 'RecoverIt' for recovery abuse and 'svc-crashcheck' for identifying vulnerable services. Furthermore, the report discusses how attackers use direct Windows API calls (e.g., CreateServiceW) to evade EDR detection of command-line tools, and the risks associated with installing kernel-mode drivers via service creation. The document concludes by emphasizing the need for a multi-layered detection strategy covering APIs, registry changes, and system events to counter these evolving threats.
Article 8
Discovery and exploitation of Bad Epoll (CVE-2026-46242), a Linux kernel use-after-free vulnerability.
TL;DR
The article details the discovery of CVE-2026-46242, a critical race-condition bug in the Linux kernel's epoll subsystem. This vulnerability enables local privilege escalation to root on various platforms including Android.
The author introduces 'Bad Epoll' (CVE-2026-46242), a significant security vulnerability identified within the Linux kernel's epoll subsystem. The flaw is characterized as a race-condition leading to a use-after-free (UAF) state. This specific vulnerability was discovered and submitted as a 0-day to the Google kernelCTF program. From a security impact perspective, the bug is highly critical because it allows an unprivileged user or process to achieve full root privileges. The scope of this threat is broad, affecting not only standard Linux server environments and desktop distributions but also the Android ecosystem. Notably, the author points out that while Anthropic's Mythos research tool successfully identified other race conditions within the same epoll codebase, it failed to detect this specific vulnerability, highlighting potential limitations in current automated security reasoning tools.
Article 9
Blockchain security experts are warning crypto protocols to implement continuous reauditing of smart contracts due to the rising efficacy of AI-driven vulnerability discovery.
TL;DR
The emergence of advanced AI tooling is enabling hackers to efficiently scan and exploit old smart contract vulnerabilities. Experts suggest that continuous reauditing is now a critical operational requirement for maintaining blockchain security.
Blockchain security landscape is undergoing a significant shift as artificial intelligence makes automated vulnerability discovery much more accessible to malicious actors. Recent reports from CertiK and TRM Labs highlight that hackers are no longer just targeting new protocols, but are actively revisiting legacy codebases to find latent bugs. A notable example includes the discovery of a major vulnerability in Zcash's Orchard pool by a security engineer using an AI agent powered by Anthropic’s Claude Opus 4.8; this bug had remained undetected for four years and posed a risk of undetectable counterfeiting. The scale of the threat is evidenced by the $1.32 billion stolen in the first half of 2026, with attackers utilizing sophisticated automated tools to scan large amounts of code at scale. Furthermore, even defunct protocols like Aztec Connect have been successfully exploited for millions of dollars. While technical audits are essential, experts also emphasize that security must extend beyond the codebase to include the disruption of broader criminal infrastructures, such as money laundering networks and state-sponsored hacking groups from North Korea. As the total value locked in DeFi exceeds $72 billion, the incentive for these AI-assisted attacks remains extremely high, necessitating a move toward continuous, recurring audit cycles rather than one-time post-deployment checks.
Article 10
The introduction of Smart Contract Oriented Programming (SCOP) via the Compose layer to enable modular, reusable, and stateless on-chain infrastructure.
TL;DR
The Compose whitepaper proposes a new development paradigm called Smart Contract Oriented Programming (SCOP) to solve the inefficiencies of redundant on-chain deployments. By utilizing stateless facets and diamond architectures, it aims to transform smart contracts from isolated programs into modular, reusable systems.
The whitepaper addresses a critical inefficiency in the current blockchain ecosystem: the lack of shared, on-chain infrastructure for reusable logic. While developers frequently reuse code via libraries and templates before deployment, this practice results in the redundant redeployment of identical bytecode across multiple addresses. This duplication increases gas consumption, complicates the audit landscape, and creates significant security risks, as evidenced by studies showing that nearly 10% of similar contract pairs share the same vulnerabilities. The paper argues that because remediation is fragmented across independent deployments, the ecosystem lacks a way to patch shared logic globally.
To resolve this, the authors introduce Smart Contract Oriented Programming (SCOPE) and the Compose layer. Unlike traditional software engineering where applications can be easily patched or replaced by their operators, smart contracts are persistent, public, and constrained by the Ethereum Virtual Machine (EVM). SCOP adapts software engineering principles—such as modularity and abstraction—to these specific on-chain constraints. The core of this approach is the separation of logic from state. By using a diamond-native architecture, Compose utilizes 'diamonds' as application containers and 'stateless facets' as reusable logic components that operate over a shared system state. This transition from contract-level to system-level reasoning allows for a more scalable, auditable, and interconnected on-chain ecosystem, effectively acting as a composition layer for decentralized applications.
Article 11
A critical Linux KVM vulnerability named Januscape allows guest virtual machines to escape to the host machine.
TL;DR
Researcher Hyunwoo Kim discovered a high-severity use-after-free vulnerability in the Linux KVM subsystem called Januscape. This flaw enables untrusted guest VMs to break out of their isolation and potentially take control of the host operating system.
A significant security flaw known as 'Januscape' has been identified within the Linux kernel's KVM (Kernel-based Virtual Machine) component. Tracked as CVE-2026-53359, this vulnerability is a use-after-free error located in the shadow MMU emulation process, which is responsible for translating memory addresses between the host and guest environments. The flaw is particularly dangerous because it can be triggered using only guest-side actions, meaning an attacker with access to a single virtual machine instance on a cloud platform could compromise the entire physical host. Specifically, the exploit allows for either a Denial of Service (DoS) by panicking the host kernel or full Remote Code Execution (RCE) with root privileges, potentially compromising all other tenant VMs sharing the same hardware. The vulnerability has remained undetected in the Linux kernel for 16 years and affects systems running on both Intel and AMD architectures. While a proof-of-concept exploit demonstrating a host crash is currently available, a full guest-to-host escape exploit has been developed but will not be released publicly for the time being. Google has recognized the importance of this discovery by awarding the researcher $250,000.
Article 12
CISA details its internal incident response following the accidental exposure of AWS GovCloud credentials in a contractor's public GitHub repository.
TL;DR
CISA released an after-action report regarding a credential leak caused by a contractor's use of a personal GitHub repository. The agency successfully mitigated the exposure without loss of mission data and is implementing stricter development environment guardrails.
CISA recently underwent an internal incident response triggered by reports from a security researcher and an investigative journalist regarding exposed AWS GovCloud keys in a public repository. Investigations revealed that a contractor had uploaded CISA build and deployment repositories, containing Infrastructure as Code (IaC) and sensitive admin credentials, to a personal GitHub account for autonomous infrastructure creation. Upon discovery, CISA's Office of the Chief Information Officer (OCIO) acted immediately to take the development environment offline, revoke the individual's access, and remove the public repository. Forensic analysis confirmed that while credentials were leaked, no customer or mission-critical data was accessed by unauthorized parties.
In the aftermath, CISA implemented several corrective measures, including a comprehensive rotation of all administrative credentials across multiple environments and the implementation of stricter controls on code repository uploads using EDR solutions. The agency's reflection on the incident emphasized the success of its Zero Trust principles and robust logging capabilities in detecting and investigating the breach. However, CISA also identified critical areas for improvement, such as the need to better monitor for secrets within private repositories, the necessity of developing specific playbooks for cloud-related incidents, and the importance of simplifying reporting channels for security researchers. Moving forward, CISA is focused on consolidating developer environments and enhancing cryptographic key agility to ensure more rapid responses to future security events.
Article 13
The debate between using ZK STARKs versus increasing block size to address the scalability and security challenges posed by post-quantum cryptography on Bitcoin.
TL;DR
The article explores the technical and governance-related dilemma of preparing Bitcoin for a quantum computing era. It compares the efficiency of using ZK STARKs for signature aggregation against the more controversial approach of increasing block size.
As quantum computing advances, Bitcoin faces a significant challenge: current digital signature schemes like ECDSA are vulnerable to quantum attacks, but replacing them with NIST-approved post-quantum alternatives results in much larger signatures. These larger signatures could drastically reduce Bitcoin's transaction capacity from thousands to just a few hundred per block. One proposed solution is the use of ZK STARKs, which can aggregate multiple large signatures into a single, small proof, potentially even increasing network throughput and preserving decentralization. However, implementing such advanced cryptography requires significant changes to Bitcoin's consensus layer, specifically through opcodes like OP_CAT or new verification instructions, which faces intense governance resistance. Another alternative is simply increasing the block size, but this 'blunt instrument' approach threatens decentralization by raising the barrier for running full nodes. While projects like Blockstream are experimenting with hash-based schemes like SHRINCS, and Ethereum is planning a post-quantum transition by 2029, the primary obstacle for Bitcoin remains its conservative governance culture rather than the underlying cryptographic capability.
Article 14
The article discusses how organizations can use a 'sniff test' approach to verify the accuracy and completeness of Software Bill of Materials (SBOMs) in hardened container images to ensure supply chain security.
TL;DR
The article advocates for a proactive 'sniff test' methodology to validate the integrity of SBOMs in containerized environments. It highlights how identifying omissions like unpinned packages or missing dependencies is crucial for preventing supply chain attacks.
As software supply chain attacks become more sophisticated, the role of the Software Bill of Materials (SBOM) has moved from a compliance checkbox to a critical security asset. Following updated CISA 2025 guidance, the article argues that an SBOM is only as valuable as its depth and accuracy, specifically regarding transitive dependencies and configuration files. The author introduces the concept of a 'sniff test'—a rapid, manual verification process that security teams can perform in minutes to detect common vulnerabilities in hardened images. These vulnerabilities include the 'latest tag trap,' where floating tags lead to unpredictable environments; missing package pinning, which prevents effective CVE patching; and the use of bloated base images that increase the attack surface. The text also addresses specific risks in specialized fields, such as AI/ML images containing deep dependency trees like PyTorch or TensorFlow. To ensure trust, the author recommends a workflow of verifying component counts, searching for essential libraries like OpenSSL or libc, and utilizing scanning tools like Syft, Trivy, or Docker Scout. Ultimately, if a vendor cannot provide a transparent, verifiable SBOM that accounts for every layer of the software stack, the integrity of the entire hardened image must be considered suspect.
Article 15
Swift has launched a blockchain-based ledger for 17 major global banks to enable 24/7 cross-border payments using tokenized deposits.
TL;DR
Swift is rolling out a new shared blockchain ledger to support 24/7 cross-border payments for 17 leading global banks. This initiative aims to integrate tokenized deposits and stablecoins into existing financial infrastructure.
Swift has officially announced that its new blockchain-based ledger is ready for initial use, with a group of 17 major global banks prepared to begin testing live transactions. The primary objective of this technological rollout is to facilitate continuous, 24/7 cross-border payment capabilities, allowing financial institutions to move funds for customers during weekends and overnight periods before final settlement occurs via traditional payment systems. The participating roster includes prominent financial entities such as UBS, BNP Paribas, BNY, Citi, HSBC, and Wells Fargo, spanning six continents. This new capability serves as a shared layer that allows banks to manage tokenized deposits—digital representations of commercial bank money—on their own respective ledgers. Crucially, Swift's approach is designed to complement rather than replace current payment rails; the platform is built to settle transactions involving stablecoins and various tokenized assets across multiple different blockchains. By bridging established financial trust with the emerging frontiers of digital currency, Swift aims to provide a unified infrastructure for the next generation of global value exchange.
Article 17
Phantom and Hyperliquid are petitioning the CFTC for regulatory exemptions for blockchain developers and non-custodial wallet providers.
TL;DR
Crypto entities Phantom and Hyperliquid have submitted a formal request to the CFTC seeking regulatory clarity for DeFi developers. They argue that blockchain protocols and non-custodial wallets should not be subject to the same regulations as traditional custodial intermediaries.
In an effort to shape the future of decentralized finance in the United States, crypto wallet provider Phantom and the Hyperliquid Policy Center have formally reached out to the Commodity Futures Trading Commission (CFTC). The core of their argument rests on the distinction between developers of open-source blockchain software and traditional financial intermediaries that hold customer assets. Specifically, the groups are requesting that the CFTC exempt blockchain protocol developers from registration requirements and ensure that non-custodial wallet providers are not legally treated as introducing brokers. Furthermore, they are advocating for a regulatory framework that allows established, regulated derivatives exchanges and clearinghouses to integrate onchain infrastructure for essential functions like trade execution, margining, and recordkeeping. This push for clarity comes amidst intense competition and legal friction between traditional finance giants and decentralized platforms. While companies like Intercontinental Exchange (ICE) have raised concerns regarding market manipulation risks in decentralized perpetual futures, other industry leaders are calling for a level playing field that permits 24/7 onchain trading. The landscape is further complicated by ongoing litigation, such as CME Group's lawsuit against the CFTC concerning its authority over crypto perpetual futures. Ultimately, the petitioners warn that without clear regulatory guidelines, American users risk being excluded from the growing onchain derivatives market as innovation shifts to offshore jurisdictions.
Article 18
Hyperliquid is leading the expansion of onchain perpetual futures into traditional financial assets like equities and commodities.
TL;DR
Pantera Capital highlights Hyperliquid's role in scaling onchain perpetual futures to include traditional assets like commodities and stocks. This shift is accompanied by growing interest from major financial institutions looking to leverage 24/7 blockchain trading capabilities.
The landscape of derivatives trading is undergoing a significant transformation as decentralized protocols move beyond cryptocurrency-native assets. According to Pantera Capital, Hyperliquid has emerged as a dominant force in this transition, capturing roughly 40% of the onchain perpetual futures market share. The structural benefits of blockchain-based perpetuals—such as continuous price discovery, no contract expiries, and 24/7 trading availability—are making them increasingly attractive to both crypto-native and traditional investors. This growth is reflected in market data showing that decentralized exchange (DEX) perpetual volumes have risen to 14% of centralized exchange (CEX) volumes since early 2023.
The expansion is not limited to crypto; there is a clear movement toward bringing equities, commodities, and indices onto the blockchain. Hyperliquid's vision involves 'housing all of finance' onchain, a goal that aligns with broader industry trends. Traditional financial giants are already engaging with these technologies: Intercontinental Exchange (ICE) has advocated for a level playing field for onchain contracts, while OKX has announced plans to launch perpetual futures based on ICE's crude oil benchmarks. Furthermore, the NYSE's partnership with Securitize and ICE's own plans for tokenized securities venues demonstrate that the integration of blockchain for 24/7 trading and instant settlement is a strategic priority for Wall Street. As these technologies mature, the boundary between decentralized finance (DeFi) and traditional finance (TradFi) continues to blur through the adoption of tokenized real-world assets.
Free weekly digest
The week’s most relevant AI, security, blockchain, and engineering stories — curated, summarised, and reviewed by humans. No spam, unsubscribe anytime.
Subscribe — it’s free