OpenAI accuses Chinese firm DeepSeek of stealing tech for chatbot
OpenAI says Chinese firm DeepSeek used "distillation techniques"—basically, shortcut methods—to train its R1 chatbot by tapping into tech built by OpenAI and other US labs.
This claim was made in a memo to the US House Select Committee on China earlier this week.
DeepSeek allegedly used obfuscated routers and hidden code
According to OpenAI, DeepSeek employees used sneaky workarounds like obfuscated routers and hidden code to dodge restrictions.
OpenAI warns that moves like this could put years of US investment at risk.
DeepSeek's new model was expected to be released around Lunar New Year 2026
DeepSeek's new model was expected to be released around Lunar New Year 2026.
While it censors some sensitive topics, OpenAI warned that safeguards may not transfer cleanly when capabilities are copied via distillation—this could increase the risk of misuse, particularly in biology and chemistry.