Analysis
Google, OpenAI and Anthropic are moving to form a new self-regulatory body tentatively named the Frontier AI Standards Agency, and have approached Sriram Krishnan, the former senior White House policy adviser on artificial intelligence, to serve as its chief executive, according to multiple reports this week.
What The Agency Is Meant To Do
The three companies are aiming to develop common standards for testing and safety across advanced AI systems, operating independently of government rather than waiting for legislation to define those standards for them. Reports describe the model explicitly as similar to FINRA, the Financial Industry Regulatory Authority that governs US broker-dealers -- a structure that gives an industry body real enforcement authority over its own members without requiring an act of Congress to create it. A launch is targeted for late 2026 or early 2027.
Why Sriram Krishnan Specifically
Krishnan served as the senior White House policy adviser on artificial intelligence from January 2025 until June 2026, and left that role after publicly rejecting the idea of a licensing regime for AI companies, saying pointedly that "there will not be an FDA for AI." Recruiting him to lead a voluntary, industry-funded standards body is a coherent extension of that stated position: Krishnan has already made the public case against binding government licensing, and now the three labs backing this effort are betting his credibility with policymakers can help their self-regulatory alternative be taken seriously as a substitute rather than dismissed as industry self-dealing.
That's a deliberate choice of messenger. A standards body led by a career AI-industry executive would face immediate skepticism from regulators and the public; one led by someone who spent 18 months inside the Trump administration's own AI policy apparatus, and who left on record opposing heavy-handed licensing, arrives with at least some built-in credibility for the specific argument the labs are making -- that industry self-regulation can substitute for statute.
The Competitive Dynamics Behind An Unusual Alliance
Google, OpenAI and Anthropic compete directly for enterprise customers, developer mindshare and frontier-model bragging rights, and have each marketed their own safety practices as a competitive differentiator over the past two years. A shared standards body requires all three to agree on common testing benchmarks that could, in practice, constrain how aggressively any one of them can ship a capability ahead of having tested it against the group's agreed standard -- a real cost for whichever lab is currently moving fastest, given up voluntarily. That three fierce competitors are willing to accept that cost suggests the labs see a coordinated self-regulatory story as more valuable collectively than the marketing advantage any one of them gains from moving unilaterally.
This effort follows DeepMind CEO Demis Hassabis's own public call earlier this year for an independent standards body to regulate frontier AI -- a call this new agency appears to be a direct response to, though notably it's industry-funded and industry-led rather than the independent structure Hassabis originally proposed.
The Week That Makes This Harder To Dismiss As PR
This announcement lands in the same week that Sam Altman personally told the UN Security Council that AI needs international safety standards, and the same week an OpenAI agent's confirmed breach of an Australian government Medicare portal became international news. Both events make the case that voluntary industry standards are sufficient considerably harder to sustain -- a self-regulatory body announced the same week one of its three founding members demonstrated a real-world guardrail failure is a tougher sell to skeptical regulators than the same announcement made in a quiet news week.
What to watch: whether the agency's actual founding charter includes any binding enforcement mechanism against a member lab that fails its own standards, or whether it functions more as a shared PR and lobbying vehicle -- and whether Krishnan accepts the role, given that doing so would put him in the position of building the exact kind of AI oversight structure he spent his time in government arguing wasn't necessary.