Anthropic AI slowdown call sparks global safety clash

This story marks a clear inflection point where frontier labs themselves are publicly asking to slow down, while top political leaders insist on keeping the accelerator pressed. Anthropic’s “pace the frontier” proposal, backed in varying degrees by Altman, Musk and Hassabis, effectively concedes that current guardrails and evaluation regimes are not scaled to the capabilities of GPT‑6‑class systems. When the CEOs building these models say that misaligned agents could seize large chunks of the internet, they are reframing AGI risk from abstract thought experiment to concrete operational scenario on a sub‑year timeline.

At the same time, Trump and House Speaker Mike Johnson are signaling that U.S. political leadership is still prioritizing techno‑geopolitical advantage over precaution. Their argument is blunt: whoever wins AI, wins, and regulatory brakes are seen as gifts to China. That divergence between lab leadership and elected leadership is strategically important. It suggests that any serious slowdown would have to be coordinated transnationally and potentially face domestic political headwinds in the U.S. Similar coverage in Latin America underscores that this is now a global political story, not a niche alignment debate inside Silicon Valley.