Slowing AI Is The Real Risk To American Leadership

Industries rarely ask to be regulated. They can stay ahead of the curve of the law and drive profits while negotiating the guardrails later. Recently, however, leaders at the center of artificial intelligence have publicly said current safeguards are not keeping pace with what AI systems can do. If they are worried, how should the rest of us feel? The answer is not to slow down. It is to get serious.

On September 12, Anthropic CEO Dario Amodei called for slowing frontier-model development, slowing the rate at which it improves its most capable models and warning that rogue agent swarms could cause internet-scale damage within months. Sam Altman and Elon Musk endorsed this proposal the same day. Amodei insists he is not asking anyone to stop training models, only to slow capability gains, but a coordinated deceleration by America’s leading labs would slow progress in effect.

On September 23, during UNGA week, Amodei and Altman brought the case to the UN Security Council, asking governments for shared testing standards and an international system for reporting serious incidents. The White House answer came in the same chamber. Science adviser Michael Kratsios told the Council that a rapidly advancing frontier is no reason to pause it and no reason to build new global governance structures around it.

A few days later in Berlin, Peter Thiel gave the answer the labs didn’t want to hear. He cautioned that he takes the risk seriously and still thinks the pause they’re asking for doesn’t exist.

While this debate is about what AI labs should do next, companies today face issues they cannot simply pause. AI Agents currently have access to internal data, software, email and even use of operational tools. Plus, Gartner predicts task-specific agents will be built into 40% of enterprise applications by the end of 2026. That is far too deep for any enterprise to just hit the brakes for months, or even weeks.