Trump science advisor says Chinese actors are copying American AI at massive scale

2026-04-23 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

The US government alleges that China is systematically copying American AI models on a massive scale, according to a memo from President Trump's science advisor, Michael Kratsios, dated April 23, 2026. Foreign actors are reportedly using tens of thousands of proxy accounts and jailbreaking techniques to extract capabilities from top US models. This process, known as distillation, creates smaller models that can match the performance of large US AI systems on specific benchmarks at a fraction of the development and operating cost. Kratsios's memo indicates that these actors also strip safety protocols from the distilled models, undermining mechanisms intended to ensure "ideologically neutral" and "truth-seeking" AI systems. The Trump administration plans to share intelligence with US companies, enhance private sector cooperation, develop joint countermeasures, and pursue action against those responsible.

Key takeaway

For CTOs and VPs of Engineering evaluating AI model security, you should prioritize understanding the specific jailbreaking and distillation techniques used by foreign actors. Implement enhanced monitoring for unauthorized access and data exfiltration, particularly concerning model chains of thought. Your teams should collaborate with government intelligence and industry peers to develop robust countermeasures against systematic IP theft and ensure the integrity of your AI systems.

Key insights

China is allegedly using large-scale distillation to copy US AI models, stripping safety features in the process.

Principles

• Legitimate distillation is valid.
• State-backed IP theft is unacceptable.

Method

Foreign actors use tens of thousands of proxy accounts and jailbreaking to distill capabilities from US AI models, creating smaller, high-performing models at reduced cost, often stripping safety protocols.

In practice

• Share intelligence on model attacks.
• Tighten private sector cooperation.

Topics

• US AI Models
• Chinese AI Copying
• Model Distillation
• AI Safety Guardrails
• Reinforcement Learning

Best for: CTO, VP of Engineering/Data, Executive, Policy Maker, AI Security Engineer, Director of AI/ML

Related on AIssential

• Anthropic calls out China's AI copycats — AI model distillation through extensive fraudulent interactions poses a significant intellectual property threat.
• US wants Claude all to itself... because it's "TOO DANGEROUS" — Advanced AI models like Claude Mythos and GPT 5.5 pose dual-use risks, prompting government control and highlighting compute scarcity.
• Links For April 2026 — Diverse topics reveal underlying principles in AI, social dynamics, and governance, often challenging conventional wisdom.
• US accuses China of “industrial-scale” AI theft. China says it’s “slander.” — The US government is preparing to counter alleged industrial-scale AI intellectual property theft by Chinese entities using distillation attacks.
• The distillation panic — Mislabeling API abuse as "distillation attacks" risks undermining a vital AI technique and fostering counterproductive regulation.
• Incriminating misaligned AI models via distillation — Distilling misaligned AI into less capable students can expose hidden misalignment by transferring propensity without evasion capability.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.