Lehane Says OpenAI Has Coordinated Safety With Rivals for Weeks
OpenAI policy chief Chris Lehane told Washington reporters that safety coordination with Anthropic and Google DeepMind is already under way and does not need an antitrust waiver.
Classy aggregates publicly available material; we did not conduct a private interview.
What changed
On 15 September 2026, Chris Lehane, OpenAI's global policy chief, told reporters in Washington that OpenAI has been coordinating on artificial intelligence safety with Anthropic and Google DeepMind for several weeks. Bloomberg first reported the briefing; TechCrunch and Reuters carried the same account. Lehane said the three frontier labs do not need an antitrust waiver to work together on safety protocols, a position that contrasts with Anthropic CEO Dario Amodei's 12 September essay, which asked Washington for a narrow waiver so rivals could slow deployment without legal exposure.
Lehane also said OpenAI supports a provision in the FRONTIER Act that would require top frontier labs to admit independent verification organizations inside their companies to audit model development. He was in Washington meeting lawmakers on catastrophic risk legislation, including bipartisan tracks associated with the Obernolte Trahan governance framework and Senate work linked to Thune, Cruz, and Klobuchar.
Why it matters
The disclosure moves three lab safety conversations from rumor to on the record policy. Enterprise buyers, insurers, and procurement teams currently rely on voluntary commitments and uneven self reporting. If coordination produces shared test categories for bioweapons uplift, cyber capability, and autonomous replication, as later reporting on the proposed Frontier AI Standards Agency described, auditors could compare labs on one checklist instead of three incompatible ones.
The waiver split matters for legal teams. Amodei argued rivals need explicit permission to coordinate slowdowns. Lehane said precedents exist in other industries, citing airline safety cooperation, and that current law already permits the talks. That gap decides whether coordination stays informal or requires new statute before labs publish joint standards.
Who is affected
Frontier lab safety and policy staff must align internal red team protocols with whatever shared framework emerges. Enterprise AI procurement teams should track whether verification org access becomes law or remains voluntary. Regulators and congressional staff now have a named OpenAI endorsement for third party lab access, which strengthens the FRONTIER Act verification language Lehane cited. Competing labs and open model developers remain outside the three company circle unless membership rules expand.
What to do next
Map your vendor risk questionnaire to the three test categories the labs are reportedly converging on: bioweapons uplift, cyber capability, and autonomous replication. Ask frontier vendors whether they participate in the coordination Lehane confirmed and whether they would accept independent verification org access if the FRONTIER Act provision passes.
What to watch
Watch for a founding charter or chief executive appointment for the proposed Frontier AI Standards Agency before year end 2026. Watch whether the FRONTIER Act verification provision survives committee markup with OpenAI's stated support. Watch whether Anthropic or Google DeepMind confirm participation on their own terms, since Lehane spoke for OpenAI only at this briefing.
Sources
- Primary. TechCrunch, OpenAI, Anthropic, Google have been in talks on AI safety for weeks (15 September 2026). On the record confirmation of multi week coordination and Lehane's waiver position.
- Secondary. Bloomberg, reporting cited by Reuters on OpenAI safety coordination with Anthropic and Google (15 September 2026). Corroborates the Washington briefing.
- Secondary. PYMNTS, OpenAI Defies Antitrust Concerns to Join Rivals on AI Safety (15 September 2026). Lehane's FRONTIER Act verification org endorsement.