U.S. Government AI Curbs Spark Market Re-Rating Fear as China Closes Gap
The Trump administration's decision to impose deployment delays on frontier models from OpenAI and Anthropic has fractured the pro-AI coalition that defined Washington's technology posture, raising the prospect of a…
HONG KONG— June 29, 2026
The Trump administration's decision to impose deployment delays on frontier models from OpenAI and Anthropic has fractured the pro-AI coalition that defined Washington's technology posture, raising the prospect of a government-controlled speed limit on American labs precisely as Chinese rivals demonstrate competitive parity on cybersecurity benchmarks.
The Policy Shift That Broke the Coalition
The White House asked OpenAI to stage the rollout of its latest model, GPT-5.6, after issuing a similar directive that forced Anthropic to suspend access to its Fable 5 and Mythos 5 models. Mythos 5 has since returned on a limited basis after Commerce Secretary Howard Lutnick stated in a letter seen by Axios that Anthropic's work with the government had "yielded significant progress." Fable 5 could follow.
The interventions drew an immediate rebuke from David Sacks, Trump's former AI and crypto czar. Writing on X, Sacks warned that restricting access to advanced American AI models risks undermining the strategy the president articulated a year ago — that the path to winning a global AI race runs through pro-innovation policy, not restriction. "We deviate from that strategy at our peril," he wrote.
China's Window Opens
The timing sharpens the strategic stakes. Two separate security evaluations show Chinese AI systems have already matched the best U.S. models on cybersecurity tasks. Open-source Chinese model usage has surged in recent weeks as developers focus on minimizing costs, with Chinese models now occupying several top positions on the OpenRouter usage leaderboard.
Box Chief Executive Aaron Levie described the development as "one of the most important changes in the AI landscape in the past four years," arguing that competitive pressure between labs has been the principal engine of rapid capability gains. Ad hoc government access decisions, rather than clear benchmarks, risk disrupting that dynamic without a compensating strategic benefit.
Investor Re-Rating Pressure
For capital markets, the signal is unambiguous. Venture capitalist Paul Kedrosky told Axios the situation is "hugely bearish," framing it as a hall monitor diluting the punch at the party — a dynamic he said would generate re-rating pressure on AI lab valuations as their most commercially valuable products become subject to government-controlled deployment timelines.
The mechanism is straightforward: if frontier model releases require official clearance, the revenue cadence and competitive moat of leading labs become harder to price with confidence, compressing multiples.
Rules vs. Ad Hoc Discretion
Not all voices oppose government involvement. Anthropic has itself urged stronger safeguards as models grow more capable. Dan Shipper, chief executive of AI subscription service Every, said the government's participation "is actually super important," provided it finds the right balance between safety and broad access.
The sharper concern is process. Mark Pincus, the Zynga founder and investor in both OpenAI and Anthropic, said he supports clear regulation but noted "it's hard to build when there's a moving target." AI startup founder Siméon Campos added a structural risk: labs could attempt to game any benchmarks used as regulatory triggers. Until rules are codified, frontier AI access is effectively being allocated by opaque government discretion — a regime investors and executives say is untenable for an industry moving at this speed.
Related reading
Source · 來源