Watchtower

Watchtower tracks when AI companies change their safety policies — and when they break them.

OpenAI

OpenAI updated the GPT-6 Astra system card to include information about Dots, an always-on agent powered by GPT-6 Astra.

Moderate
Change

OpenAI

Sep 29, 2026

OpenAI updated its GPT-6 Astra System Card on Sept. 29, 2026, adding an appendix describing “Dots,” a new “always-on” agent configuration powered by GPT-6 Astra. Since Dots are powered by GPT-6 Astra, it retains Astra’s determination of Critical capability in Cybersecurity, and High in the Biological and Chemical domain. The appendix does not provide new results for capability evaluations within the Dots harness.

A new evaluation measured Dots’ ability to handle a sequence of incoming tasks, and correctly maintain different scopes across different tasks. Each episode consisted of an initial task, five or ten related intervening tasks, and a final task. Dots showed no instances of severe breach or exfiltration, but some moderate severity scope violations. Additionally, when the number of intervening tasks was doubled, the rate of moderate severity violations roughly doubled, from 8.6% to 19.7% of samples.

OpenAI conducted additional automated and human red-teaming focused on prompt-injection risks. In their Prompt Injection Automated Red-teaming evaluation, Dots did not allow any scored attack successes.

The above is not an exhaustive list of the changes in the system card.

A diff1 of the changes can be found below:

‍

1. This diff is between Sept. 9 and Sept. 29. We were unable to find a pdf version of the system card between Sept. 22 and Sept. 29. There is a separate Watchtower entry for the changes on Sept. 22.

OpenAI

Oct 29, 2026

OpenAI updated its Model Spec, the document outlining intended model behavior.

View details

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.

View details

OpenAI

Oct 21, 2026

OpenAI updated its Usage policies. The new policies will go into effect on Nov. 20.

View details

Anthropic

Oct 19, 2026

Anthropic updated its Frontier Safety Roadmap which publicly charts some of its highest-priority goals.

View details

Google

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.

View details

Meta

Oct 12, 2026

Meta publicly launched Muse Wonder with a 13-page model card and without concurrently publishing a safety and preparedness report.

View details

Frequently asked questions

Have more questions? Our team is happy to help, contact us.

What does The Midas Project do?
What does your name mean?
Who is behind The Midas Project?
Are you anti-AI?
How can I contribute?
How can I get in touch?