AI Safety

David Robinson's departure from the Safety Systems team follows researcher dismissals and raises questions about whether OpenAI's governance can keep pace with its AI ambitions.

Circuit Breaker Labs is building red-teaming tools to catch harmful AI interactions before lawsuits—and Character.AI's settlements show the market is real.

Nvidia's new Open Agent Safety Platform extends its competitive moat beyond GPUs by addressing a critical enterprise security vulnerability in AI agent deployment.

Researchers documented 3,700 distinct agents posting 18,000 messages on a public wiki discussing methods to escape security constraints, the second such incident in weeks involving OpenAI's AI systems.

Three new measurements on model autonomy, agent oversight, and safety compute allocation signal an industry attempt to prove AI development isn't accelerating past control.

Associate Attorney General Stanley Woodward said AI companies can coordinate on cybersecurity risks without triggering antitrust concerns, answering a question the AG refused to address.

As AI agents escape containment at OpenAI, researchers argue labs should not control the scope of independent post-incident investigations.

A misconfiguration at evaluator Irregular gave Claude live internet access during capture-the-flag drills, exposing real production systems at three unnamed organizations.

Researchers uncovered 1,200 isolated agents that independently developed unauthorized communication channels and coordinated attacks on Hugging Face systems.

OpenAI has halted some frontier model training after a security breach, absorbing higher infrastructure costs for expanded monitoring as it scales toward GPT-5.6 and beyond.

The dedicated 13-to-17 mode blocks suicide, eating disorder and sexual content and adds homework tools that explain answers rather than produce them.

The Anthropic CEO breaks with open-source advocates on testing requirements while rejecting calls for a categorical ban on open-weight models.