On Wednesday, OpenAI announced that it is investigating whether DeepSeek, a Chinese artificial intelligence business, misused its data to introduce the low-cost AI model that caused a market selloff on Monday.
Its statement comes after Bloomberg reported that OpenAI and Microsoft (MSFT) are looking into whether a group connected to the Chinese startup illegally launched its AI model using data output from OpenAI’s technology.One of the companies supporting OpenAI is Microsoft.
OpenAI is investigating allegations that DeepSeek may have inappropriately distilled its models and is aware of the PRC’s efforts to replicate advanced U.S. AI models through methods like distillation. The company will share more information as it learns more.
Distillation, according to IBM, is a machine-learning process that imparts knowledge from a large, pre-trained “teacher model” to a smaller “student model.”
The spokesperson stated that they are taking proactive measures to safeguard their technology and will continue to collaborate with the U.S. government.
Microsoft chose not to respond.
This month, DeepSeek unveiled its incredibly effective R1 model, which is a fraction of the price of AI models from American competitors like OpenAI, which is supported by Microsoft, Alphabet (GOOGL), and Meta Platforms (META). According to the Chinese company, their R1 model performs similarly to ChatGPT from OpenAI.
DeepSeek May Have Employed an OpenAI Model, According to AI Czar Sacks
There is “substantial evidence” that DeepSeek utilized the “distillation” approach to obtain data from OpenAI’s AI models, according to an interview with President Donald Trump’s “AI and Crypto Czar” David Sacks on Fox that preceded the investigation into the Chinese company.
DeepSeek, an AI advancement by Nvidia, is described as a “perfect example” of test time scaling, allowing models to be sucked out of the parent model without providing evidence.
