Watchtower
Watchtower tracks when AI companies change their safety policies — and when they break them.
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
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.
The same language was replaced in both CBRN entries and Cybersecurity, “As a precautionary measure we will continue to deploy mitigations, which we assess to substantially reduce the risk posed.” was changed to “We continue to deploy mitigations.”
OpenAI
Oct 29, 2026
OpenAI updated its Model Spec, the document outlining intended model behavior.
Meta
Oct 23, 2026
Meta quietly updated Wonder’s model card nearly two weeks after the model’s release, with still no sign of a safety and preparedness report for the model.
OpenAI
Oct 21, 2026
OpenAI updated its Usage policies. The new policies will go into effect on Nov. 20.
Anthropic
Oct 19, 2026
Anthropic updated its Frontier Safety Roadmap which publicly charts some of its highest-priority goals.
Oct 13, 2026
Google launched Gemini 4 Argon along with the first “full” safety and preparedness report made under its Frontier Safety Framework (v3.1), published in April 2026.
Meta
Oct 12, 2026
Meta publicly launched Muse Wonder with a 13-page model card and without concurrently publishing a safety and preparedness report.
Frequently asked questions
Have more questions? Our team is happy to help, contact us.