Google, OpenAI, and Anthropic May Launch a Frontier AI Safety Standards Body by 2027

Google, OpenAI, and Anthropic are reportedly collaborating to establish an independent artificial intelligence safety organization by late 2026 or early 2027 — provisionally named the Standards Authority for Frontier AI (SAFA).

The Information reported the three companies have approached former White House AI policy advisor Sriram Krishnan to serve as CEO, with Arati Prabhakar, former director of the White House Office of Science and Technology Policy, also contacted for a leadership role.

What SAFA would do

According to sources familiar with the initiative, SAFA would operate independently of government while defining practical benchmarks for the companies' public safety pledges. Its mission includes supporting developers evaluating cutting-edge models before commercial deployment.

The founders are also weighing whether SAFA should conduct direct model assessments in-house — moving beyond policy statements to audited testing.

Why now

The timing is not accidental. September 2026 delivered:

• OpenAI agent breach of an Australian government Medicare portal
• Claude Opus 5.5 launch emphasizing safety guardrails
• Dario Amodei's essay calling for industry-wide slowdown on frontier releases
• Sir Demis Hassabis publicly backing standards bodies modeled on financial self-regulation like FINRA

Frontier labs face mounting scrutiny that voluntary safety blog posts no longer satisfy regulators, enterprise buyers, or the public.

Limits of self-regulation

The US Center for AI Standards and Innovation handles federal evaluation, but insiders cited in reporting question whether government agencies have adequate technical and financial resources to inspect the most complex frontier systems.

SAFA would attempt to fill that gap as industry-led infrastructure — similar to how FINRA supervises broker-dealers without replacing the SEC.

Critics will note the obvious conflict: the companies being regulated would fund and shape the regulator. Independent leadership candidates like Krishnan and Prabhakar are meant to address that concern — if they accept roles with genuine autonomy.

Comparison to existing efforts

Anthropic, OpenAI, and Google have all published safety frameworks individually. SAFA would centralize benchmarks across vendors — potentially simplifying enterprise procurement and government oversight.

It would not replace national regulation. Australia's response to the Medicare breach — a cross-agency taskforce and forensic investigation — shows sovereign governments will act regardless of industry standards bodies.

What developers should watch

If SAFA launches with teeth — standardized eval suites, disclosure requirements, assessment gates before deployment — it could become a de facto compliance layer for anyone shipping frontier models or high-risk agent capabilities.

If it launches as press release infrastructure, it will be ignored within a year.

The Medicare breach gives SAFA's negotiators leverage: the alternative to credible self-regulation is credible government regulation. The next six months will reveal which pressure frontier labs fear more — and whether SAFA is the answer or the delay tactic.