Congress Maps China's Full AI Supply Chain Offensive, and a Shoe Company Just Became an AI Stock

We covered the frontier labs' cooperation against adversarial distillation on April 14. Here is the congressional investigation that explains what they are fighting.

The House Select Committee on the Chinese Communist Party published "Buy What It Can, Steal What It Must: China's Campaign to Acquire Frontier AI Capabilities" on April 16, alongside a hearing featuring Dmitri Alperovitch, co-founder and executive chairman of Silverado Policy Accelerator. The report is the most detailed public accounting of how China acquires AI capabilities through four channels: legal procurement of chipmaking equipment (China remains the largest market despite restrictions), lawful purchase of large volumes of advanced AI chips, sophisticated smuggling networks for restricted chips, and industrial-scale fraud to extract frontier model capabilities from American AI developers. (Select Committee Report · Bloomberg Law)

The report is notable for what it recommends. It endorses six specific bills: the MATCH Act (requiring allied restrictions before unilateral controls), the AI OVERWATCH Act (replacing the current review process with affirmative export licenses for advanced AI chips), the SCALE Act (setting export limits based on China's domestic production capacity), and the Remote Access Security Act (giving BIS authority to restrict cloud access). Alperovitch's written testimony framed the stakes bluntly: "Whoever fields the best models running on the best infrastructure will likely win not just the AI race itself but the 21st Century."

The connection to agents research is direct. Distilled models carry capabilities but weaker safety constraints. When those models enter open-source ecosystems or get deployed in agentic systems, the original developer's guardrails do not transfer. The committee's finding that China extracts capabilities through "industrial-scale fraud" puts a congressional imprimatur on what Anthropic disclosed in February: 16 million Claude exchanges across roughly 24,000 fraudulent accounts.

Governance signal: The report transforms the distillation debate from a corporate security issue into a legislative agenda. If even two of the six recommended bills pass, the legal framework for AI chip and model exports changes substantially before year-end.

Investment signal: Companies whose competitive moat depends on API access to frontier models should watch the Remote Access Security Act. If BIS gains authority to restrict cloud access the same way it controls hardware exports, the enforcement surface expands significantly.