US accuses Chinese AI firms of 'malicious' copying of AI technology
Chinese AI companies secretly routed millions of user interactions through Anthropic's Claude to improve their own models, potentially exposing sensitive People's Liberation Army and police data. This follows US accusations that six Chinese firms, including Alibaba, used distillation to copy frontier models. While AI leaders like Sam Altman and Dario Amodei advocate for a coordinated development slowdown to address safety and security risks, geopolitical competition and skepticism complicate these efforts. Beijing maintains that distillation is a normal business practice and has threatened countermeasures against US crackdowns.
What changed
Reports now indicate that Chinese firms routed millions of interactions through Claude, exposing sensitive military and police data.
Live updates
-
Chinese AI Firms Allegedly Leaked PLA and Police Data to Claude
Chinese AI companies secretly routed millions of user interactions through Anthropic's Claude to improve their own models, potentially exposing sensitive People's Liberation Army and police data. This follows US accusations that six Chinese firms, including Alibaba, used distillation to copy frontier models. While AI leaders like Sam Altman and Dario Amodei advocate for a coordinated development slowdown to address safety and security risks, geopolitical competition and skepticism complicate these efforts. Beijing maintains that distillation is a normal business practice and has threatened countermeasures against US crackdowns.
Why it matters
The US views the industrial-scale theft of AI intellectual property as a national security risk. This tension exists alongside broader concerns from AI executives about superintelligence and the potential for AI to destroy humanity by 2030.
What is confirmed
- AI leaders including Amodei and Altman have called for a coordinated slowdown in AI development to address safety risks.
- Geopolitical factors and competition are hindering proposals to slow AI development.
Still unconfirmed
- Anthropic's threat intelligence report details digital espionage, kinetic weapons programs, and information warfare.
What to watch next
- Beijing's official response to the allegations of PLA and police data exposure
- US government actions or sanctions against the six named Chinese AI firms
confidence 80%Sources used for this update (6)
- www.aa.com.tr — EXPLAINER - Anthropic threat intelligence report: What to know
- www.thesaturdaypaper.com.au — Labor stands up to Trump on Digital Duty of Care
- www.rte.ie — Are AI models becoming too powerful to control?
- www.visiontimes.com — Chinese AI Firms Allegedly Exposed PLA and Police Data to Anthropic’s Claude
- enterpriseai.economictimes.indiatimes.com — Is it possible to slow down the development of AI?
- www.economy.pk — Can AI Development Really Be Slowed Down? Experts Weigh In
-
China Denies AI Theft Claims as US Names Six Firms for Model Distillation
The US government has named six Chinese AI companies, including Alibaba, for using distillation to copy frontier models like Claude and GPT. While US agencies and Anthropic describe this as industrial-scale theft of intellectual property and a national security risk, Beijing rejects these claims. China asserts that AI distillation is a normal business practice and has warned Washington that it will implement countermeasures if the US pursues a crackdown on Chinese AI firms.
Why it matters
AI distillation allows smaller models to mimic the reasoning and performance of larger, more expensive frontier models. The US is attempting to protect its technological lead by urging providers to isolate Chinese users. This dispute marks a shift from general trade tensions to specific conflicts over AI training methodologies.
What is confirmed
- The US has identified six Chinese AI firms accused of distilling top frontier models.
- Anthropic alleges that Chinese labs, including Alibaba, used mass distillation to copy the reasoning capabilities of Claude.
- Beijing describes AI distillation as a normal business practice rather than theft.
- China has warned the US of countermeasures if Washington curbs Chinese AI firms.
Still unconfirmed
- The US is calling for AI poisoning to sabotage the distillation efforts of Chinese models.
What to watch next
- Official US sanctions or restrictions targeting the six named Chinese firms
- Specific countermeasures implemented by the Chinese government
- Evidence of AI poisoning techniques being deployed by US providers
confidence 90%Sources used for this update (5)
- finance.yahoo.com — US Names Six Chinese AI Firms For Distilling America’s Top AI Models
- www.ibtimes.co.uk — China Says AI Distillation Is 'Normal' Business, Not Theft, and Warns US Against Crackdown
- asiatimes.com — US calls for AI poisoning to sabotage China’s model distillation
- www.itpro.com — What is AI distillation? The new threat facing western tech giants as US accuses Chinese firms of ‘aggressively’ copying frontier models
- www.whalesbook.com — Anthropic Accuses Chinese AI Labs of Massive Model Distillation
-
US Accuses Six Chinese AI Firms of Systematic Theft of American Models
The FBI, NSA, and CISA have accused six Chinese AI companies, including Alibaba and DeepSeek, of carrying out "industrial-scale" theft of trade secrets. US agencies claim these firms used a technique called distillation to siphon billions of tokens from American models such as GPT, Claude, Gemini, and Grok to accelerate their own development. In response, the US government is urging AI providers to identify Chinese users and secretly transition them to less-capable models to protect frontier technology. Treasury Secretary Bessent characterized the activity as stealing and copying.
Why it matters
The accusations come weeks before a leadership summit in Washington featuring a visit from Xi Jinping. This conflict centers on the strategic race for AI supremacy and the protection of intellectual property against state-sponsored industrial espionage.
What is confirmed
- The FBI, NSA, and CISA issued a joint statement accusing Chinese AI firms of systematically copying American models.
- DeepSeek and Alibaba are among the firms accused of siphoning AI models.
- Chinese firms used distillation to copy outputs from US models to speed up their own AI development.
- The US government has named six Chinese AI companies as having copied American AI secrets.
- US agencies claim Chinese firms pulled billions of tokens from GPT, Claude, Gemini, and Grok.
Still unconfirmed
- Anthropic's distillation battle has moved to the dark web.
- US authorities are urging AI firms to secretly switch Chinese users to less-capable models.
What to watch next
- The outcome of the upcoming leadership summit in Washington between US and Chinese officials.
- Official responses or denials from Alibaba and DeepSeek regarding the distillation claims.
- Implementation of restricted model access for Chinese users by US AI labs.
confidence 90%Sources used for this update (15)
- Reuters — US accuses Chinese AI firms of 'malicious' copying of AI technology
- CNN — US claims Chinese AI firms are carrying out ‘industrial-scale’ theft of trade secrets
- NBC News — U.S. agencies say top Chinese AI companies systematically copied American models
- CNBC — Anthropic's distillation battle turns to the dark web as China concerns swell
- newsquawk.com — USTR Greer and Canada's Leblanc to speak today, according to Radio Canada
- cyberscoop.com — Feds accuse China of ‘systematic’ distillation of U.S. AI models
- Bloomberg.com — US Says Alibaba, DeepSeek Have ‘Systematically’ Siphoned AI Models
- The Free Press — How China Is Stealing America’s AI
- www.nbcnews.com — U.S. agencies say top Chinese AI companies systematically copied American models
- arstechnica.com — Six Chinese AI firms accused of aggressively copying US frontier models
- www.thecourier.com.au — US accuses six Chinese AI firms of 'malicious' copying
- www.timesnownews.com — US Accuses DeepSeek, Alibaba And Other Chinese AI Firms Of Stealing AI Technology