Longtime students of this site’s columns by James Roth will remember Mr. Roth’s January 30, 2025 column about the DeepSeek miracle.
Not long after the advent of the surveilling and propagandizing CCP-bot, Roth and others discerned that “ridiculously cheap” DeepSeek, “technically equivalent to the pricey American stuff like ChatGPT,” was not all that it non-transparently seemed to careless superficial inspection!
IP theft
Roth suggested that the “whole scenario” of DeepSeek “feels like the CCP picking a winner behind the scenes…. You won’t need a large staff or a lot of money and equipment if you’re handed another Chinese company’s code…. Of course, you could also steal the work developed by your competition at great cost. IP theft from Open AI seems to be what’s happening. ‘DeepSeek’s research even admits its R1 model is based on other open-source systems.’ ”
Now we’ve got a memo from OpenAI, maker of ChatGPT, about how “Chinese artificial intelligence startup DeepSeek is targeting the ChatGPT maker and the nation’s leading AI companies to replicate models and use them for its own training.”
Reuters reports (February 12, 2026):
Sam Altman-led OpenAI accused DeepSeek of “ongoing efforts to free-ride on the capabilities developed by OpenAI and other U.S. frontier labs.”…
In the memo sent to the U.S. House Select Committee on Strategic Competition between the U.S. and the Chinese Communist Party on Thursday, OpenAI said: “We have observed accounts associated with DeepSeek employees developing methods to circumvent OpenAI’s access restrictions and access models through obfuscated third-party routers and other ways that mask their source….
“We also know that DeepSeek employees developed code to access U.S. AI models and obtain outputs for distillation in programmatic ways,” the memo added….
Hangzhou-based DeepSeek shook markets early last year with a set of AI models that rivaled some of the best offerings from the U.S., fuelling concerns in Washington that China could catch up in the AI race despite restrictions.
OpenAI said that Chinese large language models are “actively cutting corners when it comes to safely training and deploying new models.”
When I asked ChatGPT how James Roth had been able to beat OpenAI to the punch on this, ChatGPT explained that White House AI Czar David Sacks is “often misremembered by some as ‘James Roth’ ” and that it was Sacks who suggested a year ago that DeepSeek had been copying OpenAI. (My contention is that more than one person had the same thought.) The just-released OpenAI memo took time to put together and was the product of an extensive investigation, ChatGPT explained.
Also see:
OpenAI: Memo to House Select Committee on the CCP on “Updated Stakes for American-Led, Democratic AI” (February 12, 2026)
“In this memo, we detail activity we’ve observed on our platform that is indicative of ongoing attempts by DeepSeek to distill frontier models of OpenAI and other US frontier labs, including through new, obfuscated methods; how DeepSeek reflects CCP censorship and control of information in its responses, including examples; how China is providing significant state support for its frontier labs and the underlying energy needed to scale compute; steps we have taken in the last year to address adversarial distillation, and areas where further partnership with the US government would be beneficial.”
