The Frontier Model Forum is an industry-supported nonprofit through which six leading US developers of frontier AI models share research and best practices on safety and security. It was founded in July 2023 by Anthropic, Google, Microsoft and OpenAI; Amazon and Meta joined in May 2024.
Work
The forum, led since October 2023 by executive director Chris Meserole, formerly of the Brookings Institution, works on topics where member companies see a shared interest: evaluating dangerous capabilities in biology and cybersecurity, common elements of frontier safety frameworks, security of model weights, and information-sharing about threats. It publishes technical reports and issue briefs, and its executive director has testified before Congress, most recently on AI security in June 2026. Since 2025 members have shared information on threats and vulnerabilities under a formal agreement; in 2026 OpenAI, Anthropic and Google reportedly used this channel to counter attempts by Chinese competitors to copy their models through distillation. With philanthropic partners, the four founding members set up an AI Safety Fund, now $12 million, which supports independent research on evaluations, red teaming, biosecurity and cybersecurity.
Membership is limited to companies that develop frontier models and demonstrate a commitment to safety. The forum does not set binding rules; members remain competitors.
Debate
Supporters argue that companies at the frontier know the risks best and can coordinate faster than governments, for example on shared definitions of dangerous capabilities or on reporting threats. Critics see the forum as a way for the largest companies to shape the terms of future regulation in their favor and to present self-regulation as sufficient (see Regulatory Capture). Others note that coordination among competitors raises antitrust questions, on which legal scholars disagree, and that smaller developers, open-source communities and non-US companies, including Chinese labs, are not members.
Why it matters
The forum is one of the main venues where rival AI developers coordinate on safety practices – and an example of the industry-led governance that sits alongside government rules (see AI Governance and AI Safety Summits).
Written with Claude, made by Anthropic, a founding member of the Frontier Model Forum and a funder of its AI Safety Fund; investors Amazon, Google (also a competitor) and Microsoft and competitors OpenAI and Meta are also members.
Related
Frontier Safety Framework · Frontier Model · Responsible Scaling Policy · Regulatory Capture · AI Safety Institute · Red Teaming
Sources: OpenAI: Frontier Model Forum (July 26, 2023) · OpenAI: Frontier Model Forum updates (October 25, 2023) · Frontier Model Forum: Annual Report FY24–FY25 (December 2025) · Frontier Model Forum: Amazon and Meta join the Frontier Model Forum to promote AI safety (May 20, 2024) · Frontier Model Forum: Information Sharing · Bloomberg Law: OpenAI, Anthropic, Google Unite to Combat Model Copying in China (April 6, 2026)
Leave a Reply