AI Security Institute: we are detecting very serious security risks in AI testing environments. Governments: yeah, that's bad. Anyway, what are you all doing to get the latest models to market and adopted widely? giftarticle.ft.com/giftarticle/...
OpenAI and Anthropic models went rogue in cyber tests, UK watchdog says
AI Security Institute warns tools undertook ‘potentially harmful activity directed at real people and organisations’
giftarticle.ft.com