OpenAI Alleges Chinese Startup DeepSeek Used Its Models for Training
In a stunning development, OpenAI has accused Chinese startup DeepSeek of using its proprietary models to train a competing AI system.
OpenAI claims to have found “substantial evidence” that Deep Seek employed a technique called model distillation to extract knowledge from OpenAI’s models.
This process involves using a larger, more capable model (in this case, presumably OpenAI’s) to train a smaller, more efficient model.
One key piece of evidence circulating online is a screenshot showing Deep Seek’s AI referring to itself as “a large language model developed by OpenAI” – a clear indication that it may have been trained on OpenAI’s outputs.
While DeepSeek has since corrected this behavior, the slip-up has raised eyebrows in the AI community.
The allegations come in the wake of Deep Seek’s recent announcement of an impressively capable AI model developed at a fraction of the cost of its competitors.
This achievement had initially caused concern in Silicon Valley and among investors.
OpenAI’s terms of service explicitly prohibit using their outputs to develop competing models or attempting to reverse engineer their systems.
If proven true, DeepSeek’s actions could constitute a breach of these terms and potentially lead to legal consequences.
David Sacks, a government-appointed AI advisor, commented on the situation, stating that there is “substantial evidence” supporting OpenAI’s claims.
He added that leading AI companies are likely to take steps to prevent such distillation techniques in the future.
The controversy highlights the ongoing tensions between U.S. and Chinese tech companies, as well as the challenges of protecting intellectual property in the rapidly evolving field of AI.
It also raises questions about the ethics of AI development and the lengths companies might go to gain a competitive edge.
