OpenAI accuses DeepSeek of using distillation to copy US AI models and train rival systems

OpenAI Warns US Lawmakers About DeepSeek's Distillation Tactics

OpenAI has escalated its concerns about Chinese AI startup DeepSeek to US lawmakers, alleging that the company is using increasingly sophisticated methods to replicate American AI models. In a memo sent Thursday to the House Select Committee on Strategic Competition between the US and the Chinese Communist Party, OpenAI accused DeepSeek of "ongoing efforts to free-ride on the capabilities developed by OpenAI and other US frontier labs" through a process known as distillation 1

1

. The distillation technique involves using outputs from an established, powerful AI model to train a newer system, effectively transferring the older model's learnings 3

3

.

DeepSeek Employees Allegedly Circumvent Access Restrictions

The memo reveals that OpenAI has detected "new, obfuscated methods" designed to evade its defenses against misuse. According to the company, accounts associated with DeepSeek employees have been developing methods to circumvent OpenAI's access restrictions and access models through obfuscated third-party routers and other ways that mask their source 2

2

. OpenAI also stated that DeepSeek employees developed code to access U.S. AI models and obtain outputs for distillation in programmatic ways, allowing them to extract model outputs at scale 5

5

. The company points to networks of unauthorized resellers of OpenAI's services, also designed to evade controls 1

1

.

National Security Risks and Business Threats Emerge

The practice of stealing AI technology poses significant national security risks beyond commercial concerns. OpenAI highlighted that DeepSeek's chatbot had censored results about topics considered controversial by the Chinese government such as Taiwan and Tiananmen Square 1

1

. When AI capabilities are copied through distillation, safeguards often fall to the wayside, enabling more widespread misuse of AI models in high-risk areas like biology or chemistry. Since DeepSeek and many other Chinese models don't carry a monthly subscription cost, the prevalence of distillation could pose a business threat to American companies such as OpenAI and Anthropic that have invested billions of dollars in AI infrastructure and charge a fee for their premium services 1

1

.

US-China Tech Rivalry Intensifies Over AI Race

Representative John Moolenaar, the Republican chair of the House China committee, responded sharply to OpenAI's allegations, stating "this is part of the CCP's playbook: steal, copy, and kill," referring to the Chinese Communist Party 1

1

. The accusations intensify the US-China tech rivalry as Washington remains concerned about China's progress in artificial intelligence despite export controls on advanced chips. White House AI Czar David Sacks has previously warned about Chinese distillation tactics, telling Fox News that DeepSeek was "squeezing more juice" out of older chips while citing "substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI's models" 1

1

. This imbalance risks eroding the US advantage over China in the AI race, particularly as concerns mount in Washington that access to advanced AI chips may accelerate DeepSeek's progress 1

1

. OpenAI said it proactively removes users who appear to be attempting to distill its models to develop rival's models, though the memo suggests these efforts have failed to eliminate the problem 3

3

. DeepSeek and its parent company High-Flyer, along with spokespeople for the Chinese embassy in Washington, did not immediately respond to requests for comment 1

1

3

3

.