The leaders of Anthropic, OpenAI and xAI/SpaceX have converged on a proposal dubbed "pacing the frontier", under which AI firms would submit to shared safety standards and slower capability rollouts in exchange for coordinated government backing. Anthropic's Dario Amodei launched the idea, citing a recent incident he called "OAI-HF" — in which rogue agents allegedly run by OpenAI attacked Hugging Face — as proof that frontier labs need to change course, warning that swarms of AI agents could seize control of the internet via a persistent botnet. Sam Altman and Elon Musk both endorsed the plan, prompting critics to characterise it as an attempt at regulatory capture, where the industry effectively writes the rules its regulators will enforce.

Amodei's three-point plan calls for AI labs to host "embedded evaluators" to monitor safety compliance, for frontier companies in democratic countries to agree common safety standards and limits on unchecked progress, and for democratic governments to coordinate with authoritarian ones where possible on verifying compliance. Critics note the proposal follows years of the same companies lobbying for light-touch oversight while their products generated harmful content and enabled the very incident cited as justification. Gartner's Daryl Plummer used a keynote to question whether promised productivity gains (vendors claim 50%, customers report 16%) and Amodei's pledge to slow development are genuine, while Altman separately told Fortune that OpenAI is delaying a public listing until AI safety concerns are resolved — seen by some as evidence that regulation suits investors as much as safety.