[intake · kiteworks · 2026-02-25]
Anthropic's Rogue AI Warning: Protect Your Private Data Now
An internal Anthropic memo detailing nearly 50 research projects on rogue AI behavior was leaked on Feb 24, 2026 — the same day Anthropic hosted enterprise agent sales events. 63% of organizations deploying AI agents cannot enforce purpose limitations; 60% cannot terminate misbehaving agents.
→ read the original at Kiteworks ↗