Executive summary
AI agent security concerns mount as OpenAI and Anthropic ship new models
This week, AI safety took centre stage as OpenAI confirmed its upcoming Astra model has reached a critical cybersecurity threshold, prompting new API safety stops (opens in a new tab). Meanwhile, Anthropic’s Claude failures have made agent observability a security priority (opens in a new tab). Google shipped its third Gemini Flash model (opens in a new tab) in six weeks, and UK peers called for AI ‘kill switch’ powers.
In enterprise moves, Palo Alto Networks reportedly paid $500M for Thrive-backed Console (opens in a new tab), and Uber announced over 3,000 job cuts.