OpenAI says DeepSeek stole ChatGPT data sets to train its AI Model, claims to have 'solid evidence'

FirstPost ^ | 01/30/2025

[SeekAndFind]

OpenAI has claimed it found evidence suggesting that DeepSeek used distillation, a technique that extracts data from larger models to train smaller ones. OpenAI’s GPT-4 model, which cost over $100 million to train, is an example of a large and complex AI system.

OpenAI has raised serious concerns about Chinese AI startup DeepSeek, suspecting the company of using its data to train its own models. DeepSeek has gained significant attention for its cost-effective AI solutions, which are seen as strong competitors to OpenAI’s offerings. Following this, OpenAI and its partner Microsoft are now investigating whether DeepSeek used OpenAI’s API to integrate its models into their own systems.

According to sources cited by Bloomberg, Microsoft’s security researchers discovered large amounts of data being exfiltrated from OpenAI developer accounts in late 2024, which they believe are linked to DeepSeek.

OpenAI has claimed it found evidence suggesting that DeepSeek used distillation, a technique that extracts data from larger models to train smaller ones. This method is efficient, but OpenAI argues that using it to create competing models is a violation of its terms of service.

Is The distillation technique: A common practice or IP theft?

Distillation is a well-known technique in AI development, allowing smaller models to replicate the performance of more powerful ones at a fraction of the cost. OpenAI’s GPT-4 model, which cost over $100 million to train, is an example of a large and complex AI system.

However, OpenAI claims that DeepSeek has used its models to train its own system through distillation, which it argues is a violation of its terms of service. The company has not disclosed specifics of the evidence it has gathered but says it is confident that DeepSeek has used its data without permission.

[SmokingJoe]

Joke post?
What did you have to say about Biden’s use of the DOJ/FBI to wage war on his political opponents including Donald Trump for 4 years?

[Gene Eric]

the way of the dragon...

[Secret Agent Man]

Say it properly

Arr your data berong to ahhss

They don’t do “L’s”, c’mon... :)

[Secret Agent Man]

They also catch fire at a much higher rate than ours.

[SmokingJoe]

Nope.
BYD make some of the best EVs on the planet.

[agere_contra]

Agreed!

[Tell It Right]

Agreed. We hear all the time that Asians are smarter. Yet China can’t make their own IP even though they have 4 times as many brains as we do.

[teevolt]

Absolutely, oh poor babies, the whole internet business model is selling data with no compensation to the producers of it. OpenAI can pound sand

[cp124]

Figured as much. China doesn’t create, they copy and steal.

[cp124]

Hard work stealing and copying

[SmokingJoe]

Hard work producing one third of the world.s engineers every year then working super hard to innovate and create. They have just out innovated ChatGPT even coming from behind and GPT is now whining.

[Wuli]

Stealing other’s intellectual property is what the CCP has had their companies doing all along. Every western company that has set up production of their products in China has had intellectual property of theirs stolen by the CCP and then used to help Chinese startups get a “leg up”.

[Scrambler Bob]

Oh Boy, Cat Fight

[SmokingJoe]

JASON: SAM ALTMAN GOT A TASTE OF HIS OWN MEDICINE

“I think the best part of this is the fact that Sam Altman was supposed to be doing open source.

He made it a closed source company. He stole everybody's data and got caught red handed.

Now, the Chinese have come and open sourced all the stuff he stole.

I have zero sympathy for him.”

Source: @Jason, @TheAllInPod

https://x.com/MarioNawfal/status/1885534791077912997?t=d-c42xby-qb_op2E8boNYQ&s=19