Deepseek may have the world with his technological prowess, he is not as clean as he would like to make him believe. Behind his model of R-1, praised as an economic rival of Chatgpt, hides a story of massive data siphoning. Microsoft and Openai discovered a large -scale suspicious extraction in 2024 via the Openai API, and suspicions point to Deepseek. A scandal that reveals the growing tensions between China and the United States on AI.
Deepseek: Chinese AI in Microsoft viewfinder
In January 2025, the Chinese startup Deepseek, recently praised by Sam Altman, struck hard with his R-1 model, A chatgpt rival at reduced cost. But Microsoft, Number One investor in Openai, has flared a twisted blow. According to Bloomberghis safety researchers have spotted a abnormal data extraction via the Openai API in the fall of 2024.
Did the question that burns all lips: has Deepseek pumped the OpenAi models to train its own system?
David Sacks, the “tsar” from AI to the White House, has not mince his words:
“” There is substantial evidence that Deepseek distilled the knowledge of OpenAi models, and I doubt that Optai appreciates that. »»
A heavy accusation, which is based on a method called distillation: one model IA in copying another by learning from its outings.
Crispy detail: This revelation has banged the tech scholarships, erasing billions of dollars in the United States. Deepseek has won a place on the world's chessboardbut not without having some enemies.
IA and technological cold war: a merciless battle
This scandal highlights A growing confrontation Between the superpowers of the AI. OPENAI, supported by Microsoft, is on the front line of a technological war with China. And this time, the enemy is not hidden in the laboratories, but in the lines of code.
Tension is such that the US Navy has prohibited its staff from using Deepseek, fearing a security flaw. An internal email of January 24 has put things in light:
“” As a precaution, Deepseek is banned. Too high ethical and safety risks. »»
A radical measure which speaks volumes about fears around this Chinese AI.
Another concern: even if Deepseek has been designed at a lower cost, it is still based on huge data centers. As Tsarathustra reminds us:
“” American companies will learn Deepseek efficiency techniques, but data centers remain the master asset. »»
One way of saying that, despite everything, the Americans keep one step ahead.
Deepseek impresses certainly, but her meteoric ascent has a setback: she triggered a wave of crypto scams, proving that AI can also be a formidable tool for crooks.
Maximize your Cointribne experience with our 'Read to Earn' program! For each article you read, earn points and access exclusive rewards. Sign up now and start accumulating advantages.
