DeepSeek accused of using OpenAI models to train AI system
OpenAI alleges that DeepSeek improperly used distillation techniques to extract knowledge from OpenAI's proprietary models through API access to train its own competing AI system. The allegations involve potential violations of OpenAI's terms of service and raise broader questions about protecting proprietary AI models and their training data.
What happened
OpenAI alleges that DeepSeek used distillation, a technique of extracting knowledge from larger AI models through API access, to train its competing AI system without authorization.
OpenAI stated it has evidence to support the case, noting that in August 2024, OpenAI and Microsoft investigated and blocked accounts they now believe were associated with DeepSeek for suspected terms of service violations.
Reported impact
The controversy raises questions about the protection of proprietary AI models and the data used to train them, and challenges the sustainability of high-cost, closed general-purpose AI models.
Index analysis
Distillation is a technique commonly used in the AI industry that involves training smaller models using outputs from larger models, though it is often prohibited in language models' terms of service.
Governance controls
OpenAI has documented its concerns and stated it would share information as it knows more, though it clarified it was not accusing DeepSeek of a security breach.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.