Anthropic policy chief urges tighter China chip export controls to preserve US AI lead

Summary

In a POLITICO interview at the Decoded Summit, Anthropic’s head of public policy Sarah Heck argued that the United States must maintain its lead in frontier AI because “you cannot do safety from second place.” She identified tightened export controls on advanced chips to China as the single most important action for the lame-duck period, while noting that both parties agree America must lead and that domestic AI-safety regulation can proceed in parallel. The short segment focused on strategic competition rather than detailed technical benchmarks or domestic governance mechanics.

Heck spoke on the record with no anonymous sources or on-screen graphics. The framing presented export controls as an effective, straightforward tool and portrayed US leadership as a prerequisite for responsible safety work, without exploring counterarguments about enforcement efficacy or China’s progress via domestic chips and efficiency improvements.

Editorial Assessment

The broadcast accurately captures Heck’s stated views and the bipartisan consensus on US AI primacy. Viewers receive a clear statement of the “lead or lose safety capacity” thesis that Anthropic has advanced in policy circles. However, the segment omits critical recent context: the Trump administration loosened H200 export rules in early 2026, shipments described as “trivial” have occurred under license, and smuggling plus third-country cloud access have allowed China to acquire substantial compute. Chinese models now trail top US systems by only a few months on many benchmarks, and algorithmic efficiency gains have reduced the importance of raw chip volume. The presentation therefore risks leaving viewers with an overstated sense of US leverage and an underspecified view of how much safety benefit export controls actually deliver. Domestic regulation is mentioned only in passing, with no discussion of specific proposals or industry-government tensions.