\n<\/p>\n
Just a few hours after David Sacks claimed<\/a> DeepSeek used OpenAI\u2019s models to train its own models, Bloomberg Law reports<\/a> that Microsoft is investigating DeepSeek\u2019s use of OpenAI\u2019s application programming interface (API).<\/p>\n According to security researchers working for Microsoft, the Chinese company behind the R1 reasoning model<\/a> may have exfiltrated a large amount of data using OpenAI\u2019s API in the fall of 2024. Microsoft, which also happens to be OpenAI\u2019s largest shareholder, notified OpenAI of the suspicious activity.<\/p>\n While anyone can sign up and access OpenAI\u2019s API, the company\u2019s terms of service stipulate that you can\u2019t use the output to train a new AI model.<\/p>\n \u201cYou are prohibited from [\u2026] using Output to develop models that compete with OpenAI,\u201d the company writes in its terms of use<\/a>. Additionally, the company says that you can\u2019t \u201cautomatically or programmatically [extract] data or Output.\u201d<\/p>\n The core issue seems to come from distillation, a method used by AI model developers to extract knowledge from another model using a teacher-student sort of behavior. It\u2019s going to be interesting to see whether DeepSeek has found innovative ways to circumvent OpenAI\u2019s rate limits and query its API at scale. If that\u2019s the case, there will certainly be legal ramifications.<\/p>\n<\/div>\n