OpenAI, Anthropic sign Macron's Paris Call, but US and Google stay away
What's the deal? Ten AI labs — including OpenAI, Anthropic, Mistral, Microsoft, Hugging Face, Cohere, Black Forest Labs, Synthesia and Sakana — signed the European-led Paris Call on AI safety on November 9. Signatories promise shutdown mechanisms, outside testing of frontier models, and incident sharing. A growing list of governments has also backed the declaration, championed by French President Emmanuel Macron.
Who stayed out: The US and China declined to sign, as did Google and Meta. That absence leaves two of the biggest AI powers and two major developers outside the pact, limiting its reach.
What's the catch? The commitments are voluntary and carry no deadlines or penalties. Even signatory CeSIA says the pact sets no shared risk thresholds. Its promises of kill switches and outside evaluation remain aspirational rather than enforceable.
Why now? The Paris Call builds on earlier efforts such as the Hiroshima process and the Seoul frontier-AI safety commitments. It adds distinctive pledges around shutdown mechanisms and incident sharing, but keeps them non-binding.
The signal: The pact reads more as a coordination signal among some leading labs and governments than a new safety regime. The contrast sharpens as the EU AI Act's obligations for general-purpose models move into enforcement — giving Europe a harder regulatory lever than any voluntary declaration.
Read more: thenextweb.com
Image credit: UK Prime Minister
