# Canada and Cohere signed a Paris pledge on AI safety. Here's what it actually commits them to

> Six governments and ten AI companies promised shared testing, clearer liability and shutdown switches. Nothing in it is binding, and Canada still has no AI-specific law.

- Published: 2026-11-11T14:53:00.000Z
- By: The Circuit Staff
- Type: News
- Topics: AI, Policy
- URL: https://thecircuit.ca/posts/2026-11-11-paris-ai-safety-declaration-canada-cohere/

Canada signed the [Joint declaration on Paris call for AI safety and security](https://www.pm.gc.ca/en/news/statements/2026/11/09/joint-declaration-paris-call-ai-safety-and-security) on Monday, the Prime Minister's Office says. So did Toronto's Cohere. Here's what's in it, who's missing, and what it changes here.

## Who signed

- **Governments (6):** Bahrain, Canada, Finland, France, Norway and the Netherlands
- **Companies (10):** Anthropic, Black Forest Labs, Cohere, Domyn, Hugging Face, Microsoft, Mistral AI, OpenAI, Sakana AI and Synthesia
- **Civil society:** groups including Yoshua Bengio's LawZero

The U.S. and Chinese governments didn't sign, and neither did Google or Meta, [BetaKit notes](https://betakit.com/government-of-canada-signs-international-declaration-on-ai-safety-and-security/). The PMO says the declaration remains open to new endorsements.

## What it commits them to

The declaration's premise is that frontier AI risks cross borders, and that "no actor, be it a State, a company or an international organization, can solely manage all these risks." Signatories commit to three things:

1. **Shared testing.** A common framework and technical standards for evaluating frontier models, focused on misuse in cyberattacks and chemical, biological, radiological and nuclear weapons, and on losing control of a model. Signatories will publish non-sensitive test results in a standard format in a public repository.
2. **Accountability.** Governments commit to legal frameworks that clarify who is liable for "large-scale damages." AI developers commit to disclosing their models' capabilities, limits and known risks, supporting independent evaluation, and keeping fail-safe mechanisms to shut down models they control.
3. **Early warning.** Rapid alert systems for serious incidents, and confidential whistleblower channels that protect against retaliation.

## What it doesn't do

It's a declaration, not a treaty. As [The Globe and Mail reported](https://www.theglobeandmail.com/business/article-canada-joins-paris-call-for-ai-safety-and-security-standards-pledge/), it doesn't say whether or how the commitments would be enforced.

That gap matters here. Ottawa's proposed Artificial Intelligence and Data Act was part of [Bill C-27](https://www.parl.ca/legisinfo/en/bill/44-1/c-27), which stalled in committee and never passed before that Parliament's session ended in January 2025. ISED's [archived page on the act](https://ised-isde.canada.ca/site/innovation-better-canada/en/artificial-intelligence-and-data-act) says that "there is no regulatory framework in Canada specific to AI." What Ottawa has instead is a voluntary code of conduct for generative AI developers, announced in September 2023.

**Why it matters:** By signing, Ottawa has promised to clarify liability for large-scale AI harm, which would need a law Canada doesn't have yet. Watch whether that shows up in legislation, and whether Cohere's commitments to testing and disclosure turn into published results.