Final week, Chinese language lab DeepSeek launched an up to date model of its R1 reasoning AI mannequin that performs effectively on various math and coding benchmarks. The corporate didn’t reveal the supply of the information it used to coach the mannequin, however some AI researchers speculate that not less than a portion got here from Google’s Gemini household of AI.
Sam Paeach, a Melbourne-based developer who creates “emotional intelligence” evaluations for AI, revealed what he claims is proof that DeepSeek’s newest mannequin was skilled on outputs from Gemini. DeepSeek’s mannequin, known as R1-0528, prefers phrases and expressions just like these Google’s Gemini 2.5 Professional favors, mentioned Paeach in an X put up.
That’s not a smoking gun. However one other developer, the pseudonymous creator of a “free speech eval” for AI known as SpeechMap, famous the DeepSeek mannequin’s traces — the “ideas” the mannequin generates as it really works towards a conclusion — “learn like Gemini traces.”
DeepSeek has been accused of coaching on information from rival AI fashions earlier than. In December, builders noticed that DeepSeek’s V3 mannequin typically recognized itself as ChatGPT, OpenAI’s AI-powered chatbot platform, suggesting that it might’ve been skilled on ChatGPT chat logs.
Earlier this 12 months, OpenAI instructed the Monetary Occasions it discovered proof linking DeepSeek to the usage of distillation, a way to coach AI fashions by extracting information from larger, extra succesful ones. In keeping with Bloomberg, Microsoft, a detailed OpenAI collaborator and investor, detected that giant quantities of knowledge have been being exfiltrated by means of OpenAI developer accounts in late 2024 — accounts OpenAI believes are affiliated with DeepSeek.
Distillation isn’t an unusual follow, however OpenAI’s phrases of service prohibit clients from utilizing the corporate’s mannequin outputs to construct competing AI.
To be clear, many fashions misidentify themselves and converge on the identical phrases and turns of phrases. That’s as a result of the open internet, which is the place AI firms supply the majority of their coaching information, is changing into littered with AI slop. Content material farms are utilizing AI to create clickbait, and bots are flooding Reddit and X.
This “contamination,” if you’ll, has made it fairly troublesome to completely filter AI outputs from coaching datasets.
Nonetheless, AI consultants like Nathan Lambert, a researcher on the nonprofit AI analysis institute AI2, don’t suppose it’s out of the query that DeepSeek skilled on information from Google’s Gemini.
“If I used to be DeepSeek, I’d positively create a ton of artificial information from the very best API mannequin on the market,” Lambert wrote in a put up on X. “[DeepSeek is] brief on GPUs and flush with money. It’s actually successfully extra compute for them.”
Partly in an effort to stop distillation, AI firms have been ramping up safety measures.
In April, OpenAI started requiring organizations to finish an ID verification course of with a view to entry sure superior fashions. The method requires a government-issued ID from one of many international locations supported by OpenAI’s API; China isn’t on the checklist.
Elsewhere, Google not too long ago started “summarizing” the traces generated by fashions out there by means of its AI Studio developer platform, a step that makes it more difficult to coach performant rival fashions on Gemini traces. Anthropic in Could mentioned it will begin to summarize its personal mannequin’s traces, citing a necessity to guard its “aggressive benefits.”
We’ve reached out to Google for remark and can replace this piece if we hear again.
{content material}
Supply: {feed_title}
