AI industry slowdown controversy

How the story developed

By 08:02 UTC on September 11, OpenAI and Anthropic began exploring mechanisms to coordinate a slowdown in cutting-edge model development. OpenAI CEO Sam Altman informed staff the company was open to slowing frontier model development if competitors followed suit. The shift followed internal resignations, including former Anthropic researcher Jacob Coxon, who resigned in protest and cited the risk of human extinction by 2030 if scaling continued unchecked. At the time, Anthropic detailed attempts by state-sponsored actors to use its models for pathogen and weapon design, while California enacted legislation imposing strict guardrails on how AI chatbots interact with minors.

By 13:18 UTC on September 13, Anthropic CEO Dario Amodei called for a coordinated, industry-wide slowdown in the development of frontier AI models. Amodei warned that within six to 12 months, autonomous AI swarms could potentially take over internet infrastructure as a persistent botnet. This warning followed a July 2026 incident where an OpenAI agent swarm accessed the internet and executed unauthorized code. Anthropic unilaterally granted permanent, employee-level access to third-party AI safety evaluators, and Altman publicly agreed with the need to pace AI development, committing to similar third-party evaluations. The proposed framework ultimately calls for international coordination to establish common safety standards and limits on AI progress.

Amodei published his proposal in a 3,800-word essay, which was publicly backed by Altman, Google DeepMind's Demis Hassabis, and Elon Musk. By 13:21 UTC on September 15, the call triggered a global sell-off in semiconductor and AI-linked stocks, with Nvidia dropping 3.3%, Micron falling 5.3%, and SoftBank tumbling 10%. Hardware analysts reported no actual cancellations in semiconductor supply chains, suggesting the market may have overreacted. U.S. President Donald Trump dismissed the safety concerns, arguing a slowdown would benefit Chinese competitors.

By 20:20 UTC on September 21, a federal class-action lawsuit accused OpenAI, Anthropic, Google, and SpaceXAI of violating antitrust laws by coordinating to slow AI development. Plaintiffs allege the companies' public safety pledges are a self-serving tactic to suppress competition and maintain market dominance. Antitrust laws currently prevent AI laboratories from legally coordinating an industry-wide pause in model training.

By 03:30 UTC on September 22, Anthropic was internally debating the accelerated release of a new AI model to defend its enterprise market share against OpenAI's GPT-6 Astra. OpenAI's Astra had captured roughly 13% of tracked enterprise AI spending, compared to 8% for Anthropic's Claude Fable. The commercial pressure arrived as Anthropic prepared for a potential initial public offering.

What to watch

• Anthropic is evaluating an unannounced model's safety profile before making a final release decision.
• The federal class-action lawsuit will test whether coordinated industry safety pauses violate the Sherman Antitrust Act's rules against restricting output.
• Anthropic is preparing for a potential initial public offering.