When does the EU-US AI Commitments take effect?

The EU-US AI Safety Voluntary Commitments took effect on September 28, 2024.

What are the key requirements of the EU-US AI Commitments?

Pre-deployment safety testing by signatory companies Sharing safety information with governments Red-teaming for dangerous capabilities

AI Policy Desk

International·In Effect·Effective September 28, 2024

EU-US AI Safety Voluntary Commitments

Voluntary commitments by leading AI companies (OpenAI, Anthropic, Google, Microsoft, Meta, and others) to the EU and US governments, covering safety testing before model release, transparency about capabilities, red-teaming, and sharing safety information with governments. Not legally binding.

What this means for your team

These voluntary commitments apply to the AI companies that signed them, not to you as a user of their APIs. Their relevance is that they represent baseline safety practices your AI vendors have committed to — including pre-release safety testing and red-teaming. You can use these commitments as a baseline when conducting vendor due diligence: ask whether your AI vendor has signed similar commitments and what safety testing they conduct.

Key requirements

Pre-deployment safety testing by signatory companies
Sharing safety information with governments
Red-teaming for dangerous capabilities
Transparency about AI limitations and uncertainty
Voluntary — no legal obligation on signatories or their customers

EU-US AI Safety Voluntary Commitments

What this means for your team

Key requirements

Further reading

EU-US AI Safety Voluntary Commitments

What this means for your team

Key requirements

Further reading