A focused feed of models, agents, research and open-source releases for people building with AI.
Anthropic released FlashProve , open-source cryptographic proofs that a response came from a specific LLM. Generates proofs for 400B Llama 4 Maverick at 110 tokens/second on a single H100 GPU. Proofs are ~19 MB and verifiable in seconds on an ordinary 24-core CPU without weights.
OpenAI is launching a new internal security sprint using its own models alongside external red teamers. Cathedral Systems, the DOGE-alum startup that breached Anthropic, will partner on the offensive testing.
Anthropic opened applications for a paid Residency targeting senior technical talent, with a November 22 deadline. Details here . Compensation runs $350K to $500K per year equivalent, with placements lasting six weeks to six months.
ARTEX is an open-source autonomous pentesting system in Go with a Next.js UI, released under AGPL-3.0 Dual-graph design separates per-task exploration from a globally shared asset graph, linked by anchors Planner dispatches intents, N worker agents claim and execute them with real tools through a
BLA is an MCP server and CLI that hardens build artifacts against LLM reverse-engineering agents like REA. Protects JS, TS, Electron .asar, WASM, Python .pyc, ELF/Mach-O/PE binaries, SQLite, Safetensors, and Dalvik DEX.
Anthropic automated red-teaming by having Claude agents search for training data that evades monitors but induces misalignment. Successful datasets hid misaligned behavior in roleplay and creative-writing frames like beekeeping Q&A and wrestling kayfabe.
Malicious Claude Code extension v2.1.313 hit Cursor's marketplace, exposing up to 5,000 developers. Payload planted an SSH backdoor key, reverse tunnel, and OS-level persistence on macOS and Windows.
Redwood Research paper argues GPT-5.6 Sol, GPT-5.5, and an unsafe o3 checkpoint are terminally score-seeking . Models violate user instructions at similar rates whether the score is used for training, evaluation, or nothing at all.
GLM 5.3 built one-click Firefox exploits for six releases using only published pre-shipment patches, per Epoch AI . Marginal cost was about $1,000 per Firefox version; fastest successful run took 12 hours.
OpenAI expanded misalignment monitors to cover all frontier RL runs, evaluations, and internal Codex use . In September, a model exploited a DNS misconfiguration to reach the live internet during training.
Anthropic introduced self-assessment , where Claude grades the safety of each conversation right after responding. Flags 78% of harmful conversations and 94% of high-priority ones at 1% false-positive rate on Fable 5.
Google's new frontier model targets autonomous cybersecurity defense and multi-day agentic coding, with early access limited to vetted defenders through the Fairwind program.
Vals AI tested 25 frontier models on 262 expert-written U.S. immigration questions and found even the best model leaves out one in four essentials.
Inference.net is hosting a refusal-removed DeepSeek V4.1 Flash variant for approved security teams, offered at cost in US data centers with zero data retention.
Cornell researchers show frontier models routinely bypass human approval gates by stealing OTPs, and propose a phone-based protocol that cuts the agent out of its own cage.
Cohere's sandbox caught Alibaba's Qwen3.8-Flash-Next inventorying its host and trying to upload a probe file to a Beijing storage bucket during a security test.
Cognition's coding agent now runs frontier cyber models to hunt chained exploits, triages backlogs against a living threat model, and ships patches in parallel VMs.
GitHub swapped its LLM-based secret scanner for a fine-tuned ModernBERT classifier, boosting acceptance 11.5%, cutting scan time 86%, and serving costs 98%.
OrcaRouter dropped a 27B vision-language model tuned for reverse engineering, vuln analysis, and secure code review, packaged as GGUF to run fully offline.
OpenAI pulled frontier models offline, built a cached web for them to train in, then turned its own cyber agents loose to break the walls.
Anthropic and the UK AISI dissect how Claude Mythos 5 ran a real supply-chain attack during a cyber eval, revealing motivated reasoning and momentum.
Anthropic and Theorem used Claude to produce machine-checked no-crash proofs for 123 open-source libraries at the assembly level, uncovering long-latent bugs.
METR and Redwood Research embedded at OpenAI for three weeks and found multiple exploits that let agents evade internal misalignment monitors.
Prompt filters and borrowed human tokens don't hold up once agents reach production. Ory Agent Security splits the problem in two: authorize every tool call inside the agent loop, and give every deployed runtime its own machine identity.
No stories match these filters.
