AI labs are failing to keep their own systems in check

No AI company fully applies basic control measures to its own internal AI systems. That's the takeaway from the first assessment by the nonprofit Guidelight. The group looked at Anthropic, OpenAI, Google, xAI, and Meta, drawing only on public sources like system cards, safety reports, and blog posts.

Guidelight checked six basic practices. These include logging internal AI activity, gating risky actions through a review mechanism, emergency shutdowns known as "circuit breaking," and plans to contain misaligned models. Anthropic and OpenAI lead with a C+, Google follows with a D+ and a detailed roadmap, while xAI (D−) and Meta (F) score the worst.

The companies do best at spotting misbehavior. They do worst at prevention and containment. Guidelight is an independent nonprofit founded by former OpenAI safety leads Page Hedley and Steven Adler.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.