ChatGPT maker OpenAI has become the subject of widespread mockery online after whining that buzzy Chinese AI startup DeepSeek stole its intellectual property — despite indiscriminately ripping off other people’s work itself for many years.
The Sam Altman-led company told the Financial Times that it had “some evidence” of DeepSeek stealing its intellectual property, training its AI models on the output of OpenAI’s models.
DeepSeek threw Silicon Valley into complete chaos earlier this week, wiping out over $1 trillion worth of market capitalization in a single day. Its R1 model impressed with a performance that can rival OpenAI’s most sophisticated models while using a tiny fraction of the resources.
The glaring hypocrisy of OpenAI ripping off human creatives and then accusing DeepSeek of vacuuming up its work to build a better product didn’t go unnoticed.
“I’m so sorry I can’t stop laughing,” AI critic Ed Zitron wrote in a scathing post. “OpenAI, the company built on stealing literally the entire internet, is crying because DeepSeek may have trained on the outputs from ChatGPT.”
“Hahahahahahahahahahahahahahahaha hahahhahahahahahahahahahahaha,” 404 Media‘s Jason Koebler wrote in a post about the situation.
“The irony here is palpable,” another user wrote.
“You can’t steal from us!” one user joked. “We stole it fair and square!”
Other users pointed out that despite having the word “open” in its company name, OpenAI is a closed-source, for-profit entity. Meanwhile, DeepSeek’s AI models actually are open source.
“Unlike OpenAI, DeepSeek is open AI,” one X-formerly-Twitter user wrote.
OpenAI’s whining didn’t muster much sympathy.
“If you steal all the art in the whole world to make tech that lies and can’t do arithmetic, and someone steals your tech and then gives that tech out for free while you’re still charging $200/mo for it, well, maybe you’re not the actual victim here,” one Bluesky user argued.
It remains to be seen whether OpenAI has much of a case.
“DeepSeek may well have broken OpenAI’s Terms of Service and distilled their intellectual property without permission,” AI critic Gary Marcus tweeted. “OpenAI may well have done analogous things to YouTube, New York Times, and countless artists and writers.”
Marcus was referring to the New York Times‘ lawsuit against OpenAI over copyright infringement and the strong possibility that the firm trained its models by stealing videos from YouTube, which also happens to be against its terms of service.
“Karma is a bitch,” he added.
OpenAI has maintained that ripping off copyrighted material to train its AI models falls under fair use, a United States copyright law doctrine permitting the unlicensed use of protected works. Altman has also openly admitted that there would be no OpenAI if it didn’t rip off copyrighted materials.
But now that DeepSeek has allegedly run afoul of its terms of service by employing the same playbook, OpenAI’s position highlights a glaring double standard.
Meanwhile, Altman has beat the war drum, arguing that “we will obviously deliver much better models” than DeepSeek without elaborating further.
Instead of focusing on beating DeepSeek when it comes to cost efficiency, however, Altman is doubling down, arguing that OpenAI is still hellbent on delivering “more compute.”
More on DeepSeek: Another Chinese AI Company Says It’s Beaten OpenAI