Frontier Model Forum: Which AI Companies Belong and Why

Who belongs to the Frontier Model Forum, how a company qualifies, who pays, and what the group published in 2026.

Guide: AI Tools and IndustryLast verified: October 20265 min read

The Frontier Model Forum is an industry-supported nonprofit set up in 2023 by AI developers to work on frontier AI safety and security, and as of October 27, 2026 its membership page lists six companies: Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI. It publishes issue briefs and technical reports rather than products, and membership requires a financial commitment of at least three years. The group is one documented place where several large AI developers work on shared safety questions, which is a narrower and more checkable claim than any list of "top" AI companies.

Key takeaways

  • The Forum describes itself as an industry-supported nonprofit focused on frontier AI safety and security, funded by fees from member firms.
  • Members are Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI; Amazon and Meta were announced as joining on May 20, 2024.
  • Joining requires evidence of frontier-scale capability, a public safety track record, and at least three years of financial support.
  • Output consists of issue briefs and technical reports, with the latest listed brief dated September 21, 2026.
  • Membership is not a ranking, a certification, or proof that a company's models are safe.

What the Forum is and how it describes its job

The Forum's about page calls it an industry-supported nonprofit dedicated to advancing frontier AI safety and security, established in 2023. Three mandates are listed: identify best practices, advance the science of frontier AI safety and security, and facilitate information sharing. An operating board made up of representatives of member organizations oversees it. The page also says the Forum does not engage in lobbying and is funded by fees from its members.

That funding model matters when reading its output. A body paid for by the companies it studies can still publish useful technical material, but its findings come from inside the industry and are not an outside audit. Independent evaluations, government testing, and academic work remain separate sources.

Who belongs and how a company qualifies

Anthropic, Google, Microsoft, and OpenAI were the founding members. On May 20, 2024, the Forum posted that Amazon and Meta were joining them. The membership page sets out three criteria, summarized here.

CriterionWhat an applicant must show
Frontier AI capabilityEvidence of the expertise and resources to develop or deploy frontier models at scale
Safety track recordPublic acknowledgment of frontier AI risks, disclosed risk-evaluation guidelines, review processes able to delay or suspend deployment, and support for safety research including third-party evaluations
Intent to contributeParticipation in Forum workshops and activities, and financial support for its safety work for at least three years

The page does not define "frontier model" and does not publish a fee amount. Companies that are widely discussed as AI developers but do not appear on the member list, such as DeepSeek, are outside the group. The absence says nothing about their quality or safety practices, only that they are not listed members.

Timeline of documented milestones

  • 2023: the Forum is established by the four founding companies.
  • October 25, 2023: Chris Meserole, previously director of the Artificial Intelligence and Emerging Technology Initiative at the Brookings Institution, is named the first Executive Director. The same announcement introduced an AI Safety Fund with more than USD 10 million in initial funding from the four founders and philanthropic partners including the Patrick J. McGovern Foundation, the David and Lucile Packard Foundation, Eric Schmidt, and Jaan Tallinn, administered by Meridian Institute.
  • May 20, 2024: Amazon and Meta join.
  • 2026: a run of issue briefs on agents, cyber risk, and information sharing, listed below.

What it has published in 2026

The Forum's publications page lists issue briefs and technical reports. The most recent entries it shows are summarized in the table, with titles quoted as listed.

DateTypeTitle
September 21, 2026Issue briefAgents for Cyber Defense
July 31, 2026Issue briefFrontier AI Agents and Biological Tools: Preliminary Risks and Considerations
July 7, 2026Issue briefEmerging Practices for Multilingual Evaluations for CBRN and Advanced Cyber Risks
June 3, 2026Issue briefEmerging Security Practices for AI Agents
May 12, 2026Issue briefInformation Sharing, Incident Reporting, and Incident Response for Frontier AI Risks
February 13, 2026Technical reportManaging Advanced Cyber Risks in Frontier AI Frameworks

The topics show where the member companies see shared problems: evaluating models for chemical, biological, radiological, and nuclear (CBRN) and cyber misuse, securing AI agents, and reporting incidents. The documents are voluntary practice papers. They do not create legal duties, and this article did not assess their technical quality.

How to use this when reading lists of leading AI companies

  • Check whether a list states its criteria, such as revenue, model releases, funding, or research output. Lists without criteria cannot be verified.
  • Treat membership in a trade group as a fact about that group, not a quality mark.
  • Look for what a company has published and dated, and for what regulators or independent evaluators have said separately.
  • Note the date: the member list and publication list above reflect pages read on October 27, 2026 and can change.

FAQ

Is the Frontier Model Forum a regulator?

No. It is an industry-supported nonprofit that publishes voluntary best-practice papers. Rules for AI come from legislatures and regulators, not from the Forum.

Which companies are members?

The membership page lists Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI as of October 27, 2026.

Does membership mean a company's models are safe?

No. Membership criteria include a public safety track record, but the Forum does not certify or rank members' models.

Who pays for the Forum?

Its about page says it is funded by member fees, and the membership page requires financial support for at least three years. The separate AI Safety Fund drew on the founding companies and philanthropic partners.

Last verified: October 2026

Sources

  • About us, Frontier Model Forum, 2026
  • Membership, Frontier Model Forum, 2026
  • Amazon and Meta join the Frontier Model Forum to promote AI safety, Frontier Model Forum, May 20, 2024
  • Announcing Chris Meserole as Executive Director and the new AI Safety Fund, Frontier Model Forum, October 25, 2023
  • Publications, Frontier Model Forum, 2026

New to artificial intelligence?

Start with the basics, then follow the guides to tools, rules, and effects on society.

Start with AI Basics