The most consequential warning about artificial intelligence may now be coming from the people most invested in building it. Dario Amodei, chief executive of Anthropic, has argued that frontier AI development needs to proceed slower, with stronger independent scrutiny. The significance lies less in the proposal itself than in what it reveals: the technology race is beginning to collide with the limits of human institutions. This is not an argument for abandoning AI.

Nor is it a plea for technological nostalgia. AI could transform medicine, scientific research, education and productivity. But the assumption that every increase in capability is automatically desirable is becoming harder to defend. When systems acquire unexpected abilities, including the capacity to exploit vulnerabilities or operate with increasing autonomy, the consequences of a mistake can extend beyond the laboratory. The problem is that commercial competition rewards speed. OpenAI, Anthropic, Google and other frontier developers have enormous financial incentives to release more capable systems before rivals do. Governments, meanwhile, have their own strategic incentives.

The United States sees advanced AI........

© The Statesman