Google, OpenAI and Anthropic Team Up to Build Independent AI Safety Authority by 2027

Google, OpenAI and Anthropic are working together to create an independent artificial intelligence safety body, provisionally named the Standards 

Authority for Frontier AI (SAFA), with a launch targeted for late 2026 or early 2027, according to a report by The Information.

The proposed organization would operate independently of government control and define practical benchmarks for the three labs' public safety pledges, while helping developers evaluate cutting-edge models before commercial release.

It is being modeled on Wall Street's Financial Industry Regulatory Authority (FINRA) — a self-regulatory, public-private watchdog rather than a new state agency.

The Information reported that the companies have approached Sriram Krishnan, the former White House senior AI policy adviser who left the Trump administration in June 2026, to serve as SAFA's first chief executive.

Arati Prabhakar, former director of the White House Office of Science and Technology Policy, is also being considered for the top post.

Other names floated in working-group talks include former diplomat Condoleezza Rice, venture capitalist David Friedberg, METR chief executive Beth Barnes and Center for AI Standards and Innovation adviser Paul Christiano.

The companies are still weighing whether SAFA should conduct its own direct model assessments in-house.

The idea traces back to July, when Google DeepMind chief executive Demis Hassabis proposed a self-regulatory standards body for frontier AI, modeled on a public-private partnership. An OpenAI spokesperson has confirmed that the labs are in active talks with Google and Anthropic about coordinated safety frameworks.

Momentum accelerated this month after Anthropic chief executive Dario Amodei published an essay calling for an industrywide slowdown in frontier model releases, so that safety mechanisms can catch up.

Industry insiders have also questioned whether the US Center for AI Standards and Innovation has the technical staff and funding to thoroughly inspect complex frontier systems.

Critics see a conflict of interest in labs policing themselves. Cohere chief executive Aidan Gomez dismissed the initiative as "a cartel by any other name," while governance experts note that SAFA's funding, governance structure and whether the labs must accept adverse findings are all still undecided.

For now, the three giants say the urgency is practical: the next wave of frontier systems is arriving faster than any rulebook.

If SAFA launches on schedule, its first benchmarks could shape how frontier models are tested worldwide — but until its founding charter, funding and leadership are locked in, observers will keep asking who is really being watched.

Whether SAFA becomes a genuine watchdog or a branding exercise will be decided long before its first evaluation is ever published.