In April 2026, the U.S. State Department took the unusual step of sending a diplomatic cable to every American embassy and consulate on the planet with a single directive: tell your host governments that Chinese artificial intelligence companies are systematically stealing American AI intellectual property. The cable names three firms directly – DeepSeek, Moonshot AI, and MiniMax – making it one of the few times the federal government has publicly singled out specific foreign companies in a worldwide diplomatic alert.
The accusation centers on a technique called model distillation, in which a competitor queries a proprietary AI system thousands or millions of times, then uses the responses to train a cheaper copycat model that mimics the original’s capabilities. Done without permission, it can replicate years of research and billions of dollars in development costs for a fraction of the price.
The cable did not materialize out of thin air. It follows months of private-sector complaints, a bipartisan congressional investigation, and a broader U.S. campaign to slow China’s access to advanced AI technology through chip export controls and entity-list designations.
Anthropic’s complaint set the stage
Much of the technical detail behind the State Department’s warning traces back to Anthropic, the San Francisco-based AI safety company behind the Claude family of models. In February 2026, Bloomberg reported that Anthropic had accused DeepSeek and MiniMax of distilling its models for competitive gain. According to that single Bloomberg report, the company said it had identified large volumes of API queries routed through fraudulent accounts, patterns it described as consistent with automated extraction rather than ordinary usage. These specific quantitative claims have not been independently corroborated by a second outlet or an outside audit.
Anthropic also flagged Moonshot AI as a firm of concern, though public reporting has focused most heavily on DeepSeek and MiniMax. The company has not released its underlying logs for independent review, citing the need to protect proprietary security methods. That means the specific scale of the alleged copying – how many queries, over what period, and how closely the resulting models replicate Claude’s outputs – has not been verified by a neutral party.
Anthropic is not the only American AI developer to raise the alarm. OpenAI has separately alleged that Chinese entities have attempted to distill its models, a claim the company has referenced in public statements and policy filings. The overlap between these complaints gave Washington a broader evidence base than any single company’s grievance would have provided.
Congress built the political case
While Anthropic pressed its case through industry channels, the House Select Committee on the Chinese Communist Party was assembling a public dossier. The committee published a report titled “DeepSeek Unmasked,” which compiles claims about DeepSeek’s data-handling practices, security vulnerabilities, and alleged tactics for circumventing U.S. export controls on advanced chips. (The report is referenced on the committee’s website, though a direct link to the full document has not been confirmed.)
The report frames DeepSeek not as an independent startup but as a company tightly aligned with Chinese state objectives. Committee members from both parties endorsed the findings, a bipartisan posture that gave the conclusions added political weight and made it easier for the executive branch to act on them. The investigation draws on open-source research, industry input, and government assessments, interpreting all of it through a national security lens.
That adversarial mandate is worth noting. The committee exists to scrutinize entities linked to Beijing, and its reports tend to emphasize worst-case scenarios. The findings are politically significant, but readers should weigh them alongside the committee’s institutional perspective.
What the cable says – and what it does not
The existence of the diplomatic cable has been confirmed through news reporting, though the specific Reuters article that broke the story has not been independently located at a stable URL. According to those reports, the cable instructs U.S. diplomats to raise the issue of AI intellectual property theft with foreign officials and to encourage vigilance around partnerships involving the named firms. The precise wording of those instructions has not been publicly disclosed, and the paraphrased language circulating in coverage should be treated as approximate rather than verbatim.
The cable represents a concrete escalation: the U.S. government has moved from general warnings about Chinese technology risks to pointing a finger at specific companies in a document sent to every diplomatic post it operates. But the cable has not been released in full. Whether it includes classified intelligence assessments or relies solely on open-source and industry-provided data is unknown. The precise evidence it cites, the specific incidents it references, and any recommended countermeasures remain behind the diplomatic curtain.
Equally important is what is missing from the public record. None of the three named companies – DeepSeek, Moonshot AI, or MiniMax – have issued formal public responses to the cable as of late April 2026. Whether they deny the allegations, plan legal challenges, or have offered alternative explanations for the activity Anthropic described is not confirmed. No direct quotes from any U.S. official, lawmaker, or company spokesperson have appeared in the available sourcing. Without those voices on the record, the picture tilts heavily toward the accusers’ framing as filtered through secondary reporting.
Foreign governments have not shown their hand
The cable’s real-world impact depends on whether allied capitals treat it as actionable intelligence or as background noise in the broader U.S.-China technology rivalry. So far, no European or Asian government has publicly announced investigations, procurement changes, or policy reviews targeting the named firms in response to the cable. That silence could reflect the early stage of diplomatic follow-through, or it could signal that some partners prefer to navigate the issue on their own terms rather than follow Washington’s lead.
The stakes are not abstract. Governments from Paris to Tokyo to Seoul are making their own decisions about which AI providers to allow into sensitive sectors such as healthcare, defense, and critical infrastructure. A U.S. warning that specific Chinese firms are built on stolen technology could reshape those procurement calculations – or it could be dismissed as competitive maneuvering dressed up as national security policy.
Where this fits in the larger U.S.-China AI contest
The State Department cable is the latest move in a campaign that has been building for years. The Commerce Department imposed sweeping export controls on advanced AI chips in October 2022 and tightened them again in 2023. In January 2025, the Biden administration finalized an AI diffusion rule designed to limit the spread of frontier AI capabilities to adversary nations. DeepSeek’s release of its R1 model earlier in 2025, which demonstrated unexpectedly strong performance and rattled U.S. tech stocks, intensified fears that export controls alone were not enough to maintain America’s AI lead.
The diplomatic cable adds a new tool to that toolkit: international pressure. By asking allied governments to scrutinize the named firms, Washington is trying to shrink the global market available to Chinese AI companies and raise the cost of any intellectual property theft. Whether that strategy works will depend on evidence that has yet to become public, responses from the accused firms that have yet to arrive, and decisions by foreign governments that have yet to be made.
For now, the U.S. government has placed its bet clearly on the record. It believes DeepSeek, Moonshot AI, and MiniMax have benefited from the unauthorized extraction of American AI technology, and it wants the world to know it. What comes next – legal action, sanctions, trade restrictions, or simply sustained diplomatic pressure – remains an open question that will define the next phase of the global AI competition.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
