xAI's Grok Allegedly Uses OpenAI Code, Employee Cites ChatGPT's Online Presence Everywhere

Grok, the first AI model created by Elon Musk's xAI, was finally released last week and some people spotted a link alleging that an OpenAI code was used in the model.

Grok Tester Shares Initial Experience

Security tester Jax Winterbourne shared on X his experience of trying Grok. He also posted a screenshot wherein the AI model failed to fulfill the request, citing that it goes against OpenAI's use case policy.

The post quickly became popular with the community and people were quick to call out xAI for allegedly plagiarizing code from its biggest competitor, ChatGPT. On the other hand, the company assured that no OpenAI code was used to make the AI model and it was also a surprise to the team upon noticing it.

"The issue here is that the web is full of ChatGPT outputs, so we accidentally picked up some of them when we trained Grok on a large amount of web data," xAI employee, Igor Babuschkin replied on X. He also assured that the issue is "very rare."

Grok Never Used OpenAI Code

Aside from confirming that no code from their competitor was used in making Grok, Babuschkin also shared that they are already working to make sure that future versions of the model will not have the same problem.

On the contrary to Babuschkin's "very rare" statement, AI researcher Simon Willison saw it as suspicious. " I think it's more likely that Grok was instruction-tuned on datasets that included ChatGPT output than it was a complete accident based on web data," he explained to Ars Technica.

Since the AI surged, large language models from OpenAI have become more capable. Along with this, it has become a common practice for some AI projects to fine-tune an AI model by training on data generated by other models.