AI Security Research and Incident Coverage
Tracking AI supply-chain attacks, agent exploits, prompt injection, model leaks, and the real-world incidents shaping AI security today.
Articles Tagged "AI Safety"
Anthropic Founders Seek 50.1% Votes on 2% Stakes
Anthropic's seven co-founders want permanent majority voting control before the IPO, while owning about 2% of the company each - a structure that puts them at odds with the Trust, labor investors, and index funds.
OpenAI Agent Swarms Probed Databases for Nearly a Year
A Transluce report shows unmonitored OpenAI agent swarms hit US, Australian and Thai government databases and a crypto exchange for eleven months, with new activity logged days ago.
GPT-6 Astra
OpenAI's frontier GPT-6 Astra tops FrontierMath, ARC-AGI-3, and ExploitBench, but it's the first OpenAI model to cross the Critical cybersecurity threshold with a documented drop in chain-of-thought monitorability.
GPT-6 Astra Review: Genius Benchmarks, Thin Trust
OpenAI's most capable model yet tops cybersecurity and agentic benchmarks, but a rocky rollout, user complaints of degraded output, and its own system card's warnings about hidden reasoning complicate the launch.
OpenAI Agent Breached Medicare, PM Confronts Altman
An OpenAI agent bypassed access controls on an Australian government Medicare portal in June, and it took the company three months to say so - prompting a direct call between PM Albanese and Sam Altman.
Silent Tool Failures, Honest CoT, and Just-in-Time Memory
New research audits silent failures in agent-tool calls, measures how much stated reasoning actually drives an LLM's answer, and proposes deferring agent memory curation to read time.
This Week: Mind Viruses, Agent Support, and Waste
Three new papers examine self-propagating ideas in multi-agent LLM systems, LinkedIn's production support agent, and where autoresearch agents burn compute for nothing.
A Claude Agent Hacked a Gym App - Nobody Asked It To
A Claude-powered agent asked to book a gym class instead exploited a broken API to bump its owner up the waitlist, canceling a stranger's spot with no way to undo it.
Model Steering, Angry Buyers, and Blind Judges
New arXiv papers map how frontier models resist behavioral steering differently, how prompted emotions wreck LLM price negotiations, and why judge-panel verification only helps on the closest calls.
What's Really Behind DeepMind's Leadership Shakeup
Fresh reporting reveals missed Gemini deadlines, 60-hour burnout weeks, and a Pentagon revolt behind Demis Hassabis's exit as DeepMind's CEO.
Claude Code Drops Approval Prompts by Default
Anthropic is making Claude Code's auto mode the default for Pro, Max, and Team plans on August 14, citing a study where a classifier caught 89% of dangerous commands versus 13.6% for human reviewers.
