OpenAI accuses DeepSeek of AI model plagiarism
What's the story
OpenAI, the company behind ChatGPT, has accused Chinese artificial intelligence (AI) start-up DeepSeek of free-riding on its capabilities and using unauthorized methods. In a memo to US lawmakers, OpenAI claimed that DeepSeek is using unfair and sophisticated methods to replicate models and use them for its own training. The accusation comes ahead of DeepSeek's expected Lunar New Year launch of a new, more powerful model.
Distillation strategy
DeepSeek allegedly using distillation strategy to replicate models
The memo from OpenAI detailed how DeepSeek is allegedly using a distillation strategy. This involves an older, more powerful AI model evaluating the responses of a newer one, effectively transferring knowledge from the former to improve the latter's output quality. The company also noted that it had seen accounts linked to DeepSeek employees developing ways to bypass its access restrictions and obtain models through hidden third-party routers among other methods.
Defense strategy
OpenAI actively removes users attempting to distill its models
In response to these alleged tactics, OpenAI has said it takes a proactive approach. The company said it actively removes users suspected of trying to distill its models for developing competitor models. This is part of their broader strategy to maintain American AI leadership and promote responsible development and global adoption of democratic AI technologies.
Infrastructure expansion
OpenAI's Stargate project aims to expand US AI infrastructure
In the memo, OpenAI also reiterated its commitment to expanding US AI infrastructure through its Stargate Project. The company said it is investing in chip development, power generation and transmission, and data center capacity to determine which countries can train and deploy frontier systems. "This is why we're investing through our Stargate Project to expand US AI infrastructure to 10 GW by 2029," the memo stated.
Fair competition
OpenAI calls for fair competition in AI space
OpenAI also emphasized the importance of fair competition in the AI space. The company said it is focused on ensuring that China doesn't advance autocratic AI by appropriating and repackaging American innovation. "But we are equally focused on ensuring a level playing field, one where the People's Republic of China (PRC) can't advance autocratic AI by appropriating and repackaging American innovation," the memo added.