Anthropic, Google, Microsoft, and OpenAI launch the Frontier Model Forum, aiming to ensure “the safe and responsible development of frontier AI models”
Anthropic, Google, Microsoft and OpenAI team up to establish best practices but critics argue they want to avoid regulation
Financial TimesGeorge Hammond
Context & Ripple Effects
This coalition created a shared industry vehicle around frontier-model safety, but its significance is inseparable from the criticism that voluntary coordination could also shape—or preempt—government rules.
Anthropic, Google, Microsoft, and OpenAI gain a common venue to develop and promote safety practices for the most capable models.
The forum gives the companies a collective channel to engage policymakers, while immediately intensifying scrutiny of whether industry-led standards are an alternative to binding oversight.
Second-order effects
Safety procedures and definitions developed by the group can become de facto reference points for other model developers, customers, and policymakers assessing advanced-model risk.
Joint work can extend from principles to operational access controls, as the later information-sharing effort against model extraction illustrates; that raises the stakes for nonmembers whose practices differ.
Third-order effects
If such forums become durable, governance of frontier AI may increasingly be set through a small group of well-resourced labs alongside public regulation, concentrating influence over what counts as responsible deployment.
The central policy tension will persist: cooperative safeguards may improve coordination, but they can also align with member incentives to favor controlled, proprietary model access over more open alternatives.
The trend: This is an early instance of frontier AI governance shifting from individual lab pledges toward collective, industry-led institutions that can influence both safety practice and market access.
Companies creating #AI have a responsibility to ensure that it is safe, secure, and under human control. This initiative is a vital step to bring the tech sector together in advancing AI responsibly and tackling challenges so that it benefits all humanity. https://blogs.microsoft…
🤝 Who can join the FMF? - Organizations that develop and deploy frontier models, - Demonstrate strong commitment to frontier model safety, and - Are willing to contribute to advancing the Forum's efforts. 4/
Anthropic, Google, Microsoft and OpenAI are launching the Frontier Model Forum, an industry body focused on ensuring safe and responsible development of frontier AI models. Great to see more focus on AI safety specifically. https://blog.google/... [image]
Thrilled to bring @GoogleDeepMind's expertise on responsible AI to the new Frontier Model Forum and look forward to collaborating with policymakers, academics, civil society & companies to advance AI safety. https://blog.google/...
Anthropic is excited to collaborate with industry, civil society, government, and academia to promote safe and responsible AI development. The FMF will play a vital role in coordinating best practices and sharing research on frontier AI safety. https://anthropic.com/...
Today we're joining @AnthropicAI, @Microsoft and @OpenAI to announce the Frontier Model Forum, a new group promoting the responsible development of advanced AI models. https://blog.google/...
Announcing Frontier Model Forum, an industry body co-founded with @anthropicAI, @Google, @googledeepmind, and @microsoft focused on ensuring safe development of future hyperscale AI models: https://openai.com/...
While I may disagree on the likelihood of extinction scenarios from super-AI, I agree with many of the arguments made by Y. Bengio in this statement. Following healthy discussions about extreme hypothetical risks, its time to explore concrete action plans https://yoshuabengio.org…
Today, our CEO Dario Amodei has the opportunity to discuss the risks and oversight of AI in front of the Senate. You can read his full testimony here: https://www.judiciary.senate.gov/ ...
“I don't think anybody really understands yet what the rel value is going to be of the larger, more gen-purp models vs smaller, more tgt models, which will be less exp to build & train. We're trying to set up for a world in which our cust want both types” https://www.ft.com/...