OpenAI Anthropic Google Tri-Lab Safety Coordination 2026

Published Washington, D.C., September 15, 2026.

OpenAI, Anthropic, and Google DeepMind have been in active talks for several weeks about a coordinated AI safety framework, OpenAI global policy chief Chris Lehane told reporters on September 15, 2026, and the labs do not see a need for an antitrust waiver.

Quick Answer

On September 15, 2026, OpenAI global policy chief Chris Lehane confirmed that OpenAI, Anthropic, and Google DeepMind are in active talks on a coordinated AI safety framework. The talks build on Demis Hassabis's July 14, 2026 essay proposing a US-led standards body modeled on FINRA, Dario Amodei's September 13 deceleration call, and Sam Altman's September 14 public agreement with Amodei. Lehane said no antitrust waiver is needed (Bloomberg, September 15, 2026).

Last verified: Sep 15, 2026.

At a glance

• Tri-lab talks have been ongoing for several weeks, confirmed September 15, 2026
• Lehane: no antitrust waiver needed for safety coordination
• Framework builds on Hassabis July essay, Amodei deceleration, Altman agreement
• FRONTIER Act introduced in Senate on September 15, 2026
• Tri-lab joint statement expected within months, no firm date

Background to the tri-lab talks

Three published positions from the lab CEOs converged in September 2026. Demis Hassabis of Google DeepMind published a July 14, 2026 essay proposing a US-led standards body modeled on FINRA, the Financial Industry Regulatory Authority. FINRA is industry-funded, industry-governed, and has enforcement teeth to discipline non-compliant broker-dealers. Hassabis argued that the AI sector needs an equivalent for frontier AI safety (Crypto Briefing, September 14, 2026).

Dario Amodei, CEO of Anthropic, published an essay on September 13, 2026 calling for a deceleration of frontier AI development. Amodei estimated that without coordinated slowdown, frontier AI capability could exceed human oversight capacity within two to four years. Sam Altman of OpenAI publicly agreed with Amodei's position on September 14, 2026, saying he agreed with Dario that we need to pace the frontier (TechCrunch, September 15, 2026).

These three positions are not identical. Hassabis wants a standards body with industry governance. Amodei wants a slowdown. Altman agrees with the slowdown in principle but has not committed to specific limits. The tri-lab talks are an attempt to find common ground across these positions (Bloomberg, September 15, 2026).

What Lehane said on September 15

Lehane confirmed the talks in a press briefing and addressed the antitrust question directly. Lehane, OpenAI's global policy chief and a former Airbnb and Hillary Clinton campaign aide, told reporters on a September 15, 2026 call that the tri-lab talks have been ongoing for several weeks. He said the safety coordination does not involve price-fixing, market allocation, or any other conduct that would trigger antitrust concerns (Bloomberg, September 15, 2026).

Lehane added that the safety coordination is consistent with existing trade association practice in regulated industries like finance and pharma. He cited the Pharmaceutical Research and Manufacturers of America (PhRMA) and the American Bankers Association as examples of industry groups that coordinate safety without antitrust concerns. He said OpenAI wants the standards body to have third-party verification and enforcement teeth, similar to FINRA's role (Bloomberg Law, September 15, 2026).

Lehane did not commit to a specific public timeline. He said the labs would publish a joint statement when an agreement is reached but declined to say when that would be (Bloomberg, September 15, 2026).

The FRONTIER Act and Catastrophic Risk Bill

The Senate's Catastrophic Risk Bill introduced on September 15, 2026 includes FRONTIER Act provisions. Senators John Thune, Ted Cruz, and Amy Klobuchar introduced the bill on September 15, 2026, the day before the tri-lab talks were confirmed. The bill has 14 co-sponsors from both parties and is expected to move to committee markup in October 2026 (TIME, September 15, 2026).

The FRONTIER Act provisions within the bill would establish an independent verification organization for frontier AI safety claims. The organization would have subpoena power, could audit AI labs above a compute threshold, and would publish annual safety reports. The verification organization is a step toward the standards body Hassabis proposed in July, but with government backing rather than purely industry funding (Moneycontrol, September 15, 2026).

The White House has not yet taken a formal position on the Catastrophic Risk Bill. The September 14, 2026 hoax comments from Trump at the All-In Summit complicate the legislative path, but the bill's bipartisan co-sponsorship suggests the bill can move forward regardless of the executive branch position (Bloomberg, September 15, 2026).

Industry and academic reaction

Reaction to the tri-lab confirmation has been positive but cautious. Microsoft, Meta, and Apple have not joined the tri-lab talks but have signaled interest in observing the outcome. AI safety researchers at Stanford, MIT, and UC Berkeley have called for the talks to include independent academic observers, similar to the CERN model for high-energy physics (TechCrunch, September 15, 2026).

Critics, including some AI startup founders, have warned that a tri-lab standards body could become a barrier to entry dressed up in safety language. The Open Markets Institute, a competition policy think tank, has raised concerns that the standards body could function as a cartel device. The tri-lab talks have not addressed these criticisms publicly (Crypto Briefing, September 15, 2026).

The next major test of the tri-lab process comes on September 24, 2026, when the Senate Commerce Committee holds a hearing on the Catastrophic Risk Bill. Anthropic, OpenAI, and Google DeepMind have all been invited to testify, and the hearing will be the first public confirmation that the tri-lab talks have produced concrete policy positions (US Senate Commerce Committee, September 15, 2026).

Data last verified September 15, 2026 from Bloomberg, TechCrunch, Crypto Briefing, Bloomberg Law, TIME, Moneycontrol, and US Senate Commerce Committee filings.