Google, OpenAI and Anthropic push a Frontier AI standards body — Meta, xAI and Nvidia say no
Forkast, Sunday night, citing The Information: SAFA is being formalized. Lehane confirmed the talks on Sep 15. Zuckerberg and Huang opposed at Dreamforce. Hawley and Cruz blocked an NDAA exemption; Buist v. Anthropic calls safety coordination a Sherman Act cartel.
Feature2 minIn-depth
Priya Shah2 min read
- SAFA
- OpenAI
- Meta
- Nvidia
- Antitrust

Photo: Forkast NewsSource page
The three-lab safety body already has a working name on this site: SAFA, late 2026 or early 2027, still unincorporated. Lena Park’s Sunday Forkast file is a different cut. Google, OpenAI and Anthropic are, she wrote, formalizing a FINRA-inspired Standards Authority for Frontier AI — common benchmarks, model testing, auditor qualifications, incident reporting. Meta, xAI and Nvidia have said no in public. The Information’s Sep 23–24 reporting is the paywalled primary; Park and Leo Schwartz at VFF are the secondary confirmation used here. Treat the charter as a plan until someone files one.
Chris Lehane, OpenAI’s chief global affairs officer, confirmed industry-standards talks to Bloomberg, CNBC and TechCrunch on Sep 15. On Sep 9 he published an OpenAI post, “The AI policy window is open. We need to act,” arguing for national safety requirements, state momentum and industry-led standards. He has also backed a FRONTIER Act provision that would put independent verification organizations inside frontier labs. OpenAI published its own third-party assessment framework on Sep 22. That is OpenAI defining a method while asking Washington for one.
Leadership names in the secondary reporting still include Sriram Krishnan, a former White House AI adviser, and Arati Prabhakar. Park also lists Condoleezza Rice and David Friedberg as people who have been approached. Schwartz wrote that Krishnan had been approached for a chief-executive role. None of those names have been confirmed as accepted. The Sunday file on this site already logged Krishnan and Prabhakar. The new names are reported approaches, not appointments.
Dreamforce no, and a lawsuit that calls it a cartel
At Dreamforce on Sep 15, Park wrote, Meta, xAI and Nvidia opposed forming SAFA. Mark Zuckerberg’s line was that existing liability is enough. Jensen Huang told Marc Benioff that AI laws are “just completely unnecessary” and that model makers should take responsibility for their products. Those remarks sit next to Huang’s earlier Dreamforce comments already on this site; the SAFA-specific refusal is the new fact. A standards body that launches without two frontier labs and the main chip supplier is a club, not a sector rule.
The legal weather is hostile. Senators Josh Hawley and Ted Cruz blocked an NDAA antitrust exemption on Sep 15 that would have given coordinated safety standards cover. Three days later, Buist et al. v. Anthropic PBC et al. alleged that safety coordination among competitors is an output-restricting cartel under Section 1 of the Sherman Act. Cohere’s Aidan Gomez has called the proposed body “a cartel by any other name.” FINRA works because Congress gave it statute. SAFA, as described, is a voluntary pact. That is exactly what the complaint says is illegal. Until there is a charter, a board and a yes or no from Meta and xAI, it remains a reported project with named opponents.


