OpenAI's global policy chief Chris Lehane said Tuesday that the company's engagement with Anthropic and Google had been under way for several weeks. He added that OpenAI does not see the need for an antitrust waiver for the three AI firms to coordinate on safety matters. The disclosure came September 15. The chairman of the U.S. Federal Trade Commission said on Tuesday that everyone should be deeply suspicious about AI companies seeking antitrust exemptions while they also lobby for new regulations. That is the gap. OpenAI says no waiver is necessary. The FTC chair says be suspicious. The API does not surface which interpretation applies to my production traffic.
I run ten platforms that call OpenAI and Anthropic models in production. The construction ERP at coenconstruction.com uses GPT-4 to parse change orders. The estimating SaaS at estimate.pro uses Claude to extract cost data. The review classifier on valhalla-k9.com uses GPT-4o mini for sentiment. None of those systems have a fallback for "the vendor stopped shipping this capability because three labs agreed to slow development and the FTC is investigating whether that constitutes collusion." The API either returns a completion or it does not. The coordination either affects my latency or it does not.
The labs coordinated for weeks before announcing
Anthropic CEO Dario Amodei is calling for an immediate slowdown in the pace of AI development, warning of potentially devastating consequences in a matter of months otherwise; OpenAI CEO Sam Altman quickly agreed that the industry needs to slow the pace of frontier-model advances and take more steps on safety. Amodei published the call for a slowdown on September 12. Amodei wrote that over the last few months he has become convinced that fully addressing the risks requires even more prudence—not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up. That is a commitment to slow model development. It is not a commitment to tell API users when a capability gets removed because three labs agreed it was too risky to ship.
Amodei envisioned leading AI companies operating in democratic nations reaching agreement on shared safety benchmarks and constraints on how quickly capabilities can advance; Amodei acknowledged this would require government support, including antitrust waivers to allow safety-related discussions between competitors. The labs are already talking. OpenAI confirmed the discussions have been running for weeks. Lehane also argued that no antitrust waiver is necessary for OpenAI, Anthropic, and Google DeepMind to work together on safety. That is the legal question. Amodei asked for a waiver. Lehane says no waiver is needed. The two statements do not agree.
The FTC chair is suspicious. OpenAI says proceed anyway.
FTC Chairman Andrew Ferguson said he was expressing his personal opinion, and did not directly name Anthropic, whose CEO last week called for antitrust permission to enable rivals to coordinate a slower pace to AI development due to safety concerns. Ferguson spoke on September 15, the same day Lehane confirmed the coordination. FTC Chairman Andrew Ferguson expressed skepticism toward AI firms that simultaneously request new regulations and antitrust exemptions; his concern is straightforward—if the biggest AI companies get permission to coordinate on safety standards, those standards could end up being designed, intentionally or not, to lock out smaller competitors. That is not a theoretical concern. That is a statement from the FTC chair that the coordination looks like collusion.
The three labs are fierce competitors. They are coordinating on safety. They did not announce the coordination. They confirmed it when reporters asked. OpenAI confirmed it is in active talks with Anthropic and Google DeepMind to coordinate on AI safety, marking one of the most direct admissions yet that the industry's fiercest rivals are quietly building a shared framework to manage risk from frontier models; the confirmation landed on September 15, 2026. The word is quietly. The labs were not planning to disclose the talks. They disclosed them when the reporting forced the issue. I call the API every day. The API does not tell me which capabilities the three labs agreed not to ship.
I have written before about models saying AGI while the API says staged rollout and vendors drawing red lines the judge said were legal. This is the same problem. The labs coordinate at the policy level. The API surfaces the result at the request level. If the coordination produces a capability gate, the API either documents the gate or it does not. If the gate breaks my application, my error handler either catches it or it does not. The problem is not that the labs are talking. The problem is that the API does not reflect what the labs agreed to until after my application breaks.
The coordination is either legal or it is collusion. The API does not tell me which.
Lehane said OpenAI supports a provision in the FRONTIER Act that would require top AI labs to allow independent verification organizations into their companies to monitor model development; any AI bill becoming law before the end of 2026 is a long shot; the House departs Washington on Thursday and has no scheduled sessions before voters head to the polls in November. The legislation might not pass. The coordination is already happening. OpenAI's Sam Altman, Google's Demis Hassabis, and Elon Musk all expressed support for Amodei's proposal, with Altman indicating OpenAI would join Anthropic in embedding third-party evaluators to monitor safety internally. The evaluators are going in. The API is not getting a route that tells me what the evaluators flagged. The coordination continues. The FTC remains suspicious. I keep calling the API. The response either matches the SLA or it does not. The legal risk is upstream. The operational risk is mine.