Link: OpenAI has evidence that its models helped train China’s DeepSeek

Chinese AI company DeepSeek has introduced competitively priced AI models, triggering suspicions at OpenAI of using proprietary data without permission. OpenAI, in collaboration with Microsoft, is investigating possible unauthorized use of their AI models through DeepSeek's integration efforts.

According to reports by Bloomberg, Microsoft's security team noticed significant data transfers from OpenAI's developer accounts in late 2024, suspecting DeepSeek's involvement. DeepSeek allegedly utilized a technique known as distillation to train their AI models, a method considered efficient yet controversial.

While distillation helps in efficiently training AI models, it violates OpenAI’s terms of service if used to create competing products. OpenAI has discovered evidence linking DeepSeek to this practice but hasn't disclosed specifics.

David Sacks, an AI advisor to former President Donald Trump, acknowledged the possibility of intellectual property theft by DeepSeek. He emphasized the irony of the situation, given OpenAI's own history of extensively collecting web data.

OpenAI has issued statements confirming ongoing attempts by Chinese firms to replicate U.S. AI technologies. The company stresses the importance of collaboration with the U.S. government to safeguard their innovative AI models.

This case underscores the complexities of AI development and intellectual property in the global tech landscape, where companies must navigate both innovation and legal constraints. #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.