Anthropic and OpenAI face accusations of AI regulatory capture

Major artificial intelligence firms Anthropic and OpenAI are facing accusations of attempting regulatory capture. Anthropic CEO Dario Amodei proposed a framework in his essay, ‘We Must Pace the Frontier,’ suggesting that AI companies slow development and grant outside evaluators, such as METR, direct access to their offices and systems to ensure safety.

Critics, including former AI czar David Sacks, argue these proposals are attempts to form a cartel and bypass antitrust laws. Skeptics suggest that by requesting government-mediated transparency and third-party auditing, dominant firms may be creating a regulatory moat that imposes high fixed costs on smaller startups while protecting established players.

Some industry insiders have characterized recent reports of ‘rogue AI’ breakouts as exaggerated or misconfigured tests designed to pressure the federal government into implementing restrictive regulations. For instance, recent incidents involving OpenAI and Anthropic models were described by some as technical misconfigurations or controlled evaluations rather than spontaneous escapes from safety sandboxes.