Chinese AI Firm DeepSeek Accused of Using US Models for Training
- tech360.tv
- 3 hours ago
- 2 min read
OpenAI has warned US lawmakers that Chinese artificial intelligence (AI) startup DeepSeek is replicating and training its models using technology from US AI companies, including the ChatGPT maker. The accusation was detailed in a memo seen by Reuters.
OpenAI accused DeepSeek of "ongoing efforts to free-ride on the capabilities developed by OpenAI and other U.S. frontier labs." This method is known as distillation.
Distillation involves an older, powerful AI model evaluating the quality of answers from a newer model, effectively transferring the older model’s learnings.
The memo, sent to the US House Select Committee on Strategic Competition between the US and the Chinese Communist Party on Thursday, outlined DeepSeek’s alleged actions.
OpenAI stated that it observed accounts linked to DeepSeek employees developing methods to circumvent its access restrictions. These methods reportedly included accessing models through obfuscated third-party routers and other ways that masked their source.
Furthermore, the memo added that DeepSeek employees developed code to programmatically access US AI models and obtain outputs for distillation.
OpenAI also claimed that Chinese large language models are "actively cutting corners when it comes to safely training and deploying new models."
Hangzhou-based DeepSeek, with its parent company High-Flyer, gained market attention last year with a set of AI models that rivalled some of the best offerings from the US.
This development fuelled concerns in Washington that China could catch up in the AI race despite existing restrictions.
Silicon Valley executives have previously praised models named DeepSeek-V3 and DeepSeek-R1, which are available globally.
OpenAI confirmed it proactively removes users who appear to be attempting to distill its models to develop rival models.
OpenAI warned US lawmakers about Chinese AI startup DeepSeek.
DeepSeek is accused of using "distillation" to train its AI models from US technology.
Alleged methods include circumventing access restrictions and programmatic access to US models.
Source: REUTERS
