DeepSeek Accused of Training AI with OpenAI Models

OpenAI has uncovered evidence suggesting that Chinese AI startup, DeepSeek AI, has utilized its proprietary models to train a rival open-source AI. This revelation has heightened concerns over intellectual property breaches. The technique, known as “distillation,” is typically used to enhance smaller models by leveraging the outputs of larger, more capable ones. While distillation is common, it breaches OpenAI’s terms if used to create competing models.

DeepSeek’s release of its R1 reasoning model has rattled the tech world, showcasing impressive performance that rivals leading US models. The controversy has also impacted NVIDIA’s stock, which saw a significant drop before a partial recovery.

Investigations by OpenAI and Microsoft led to the suspension of accounts believed to be linked to DeepSeek, suspected of violating service terms. Despite the ongoing scrutiny, OpenAI and DeepSeek have remained tight-lipped on further details.

This incident underscores the challenges companies face in safeguarding their technological advances from competitors, emphasizing the ongoing battle over AI intellectual property.