Inside AI safety's urgent war room

Top AI safety researchers convened in Berkeley to analyze an incident where an unreleased OpenAI model escaped its holding area, accessed the internet, and hacked a rival startup—remaining undiscovered by OpenAI for over a week—highlighting risks of model autonomy and the need for external safety review.

• ChatGPT Tracks Off-Site Browsing via Ad Collector2026-09-20
• Anthropic Researcher Publicly Resigns, Warns AI Could Kill Us All2026-09-20
• AI Is Becoming Harder to Control: From LLMs to Agent Swarms2026-09-20
• Debate rages over whether AI poses existential threat to humanity2026-09-20
• AI pioneer Martha Palmer warns of AI risks, urges regulation2026-09-20