Watchtower
Watchtower tracks when AI companies change their safety policies — and when they break them.
Released a responsible scaling policy, entitled their Frontier AI Framework.
Meta
Feb 3, 2025
On February 3, 2025, Meta released its version of a responsible scaling policy, entitled their Frontier AI Framework.
The full PDF of the policy can be read here. While it generally addresses the same topics as previous policies, it does not include any specific details about mitigations, and as such, there is no connection between specific dangerous capability evaluations and the safety measures that will be taken to address them.
However, they do say that when sufficiently high risk thresholds are reached, they will halt development or deployment to address them. They also provide abstract descriptions of some catastrophic outcomes related to cyber attacks and CBRN risks that are in scope for the policy.
Anthropic
Sep 22, 2026
Anthropic’s release of Opus 5.5 with an “inconclusive” determination for harmful-manipulation.
xAI
Sep 21, 2026
xAI says Grok 4.7 scores below its safety framework’s capability thresholds on dual-use knowledge, but xAI’s safety framework doesn’t provide thresholds.
OpenAI
Sep 9, 2026
Updated the GPT-6 Astra system card, specifically to hedge claims about misalignment.
OpenAI
Aug 18, 2026
OpenAI updated its Model Spec, the document outlining intended model behavior
xAI
Aug 17, 2026
xAI updated Grok 4.6's model card after release. A changelog was included.
Aug 14, 2026
Google updated its Gemini 3.7 Flash model card after publication, making changes to language in the “Key Results for Gemini 3.7 Flash” column in the Frontier Safety Assessment section
Frequently asked questions
Have more questions? Our team is happy to help, contact us.