OpenAI claims it has found evidence that Chinese AI startup Deepseek used its models without permission to train competing products, highlighting ongoing tensions around AI development practices.
The Financial Times and Bloomberg report that Microsoft and OpenAI discovered signs that Deepseek used OpenAI's proprietary models to improve its own alternatives. OpenAI told the Financial Times it found evidence of "distillation" - a technique where developers use larger, more powerful models to enhance the performance of smaller ones.
While distillation is common in research settings, using it to develop competing products violates most providers' terms of service, including OpenAI's.
OpenAI and Microsoft blocked suspicious accounts
Microsoft's security team noticed unusual patterns last fall, when accounts possibly linked to Deepseek pulled large amounts of data through OpenAI's API. After investigating the suspicious activity, both Microsoft and OpenAI blocked the accounts, suspecting systematic violations of their terms through unauthorized distillation.
"We know PRC based companies — and others — are constantly trying to distill the models of leading US AI companies" OpenAI said in a recent statement, without directly addressing Deepseek. The company is working closely with the US government to protect the most powerful models from such attempts, it added.
