In a TechCrunch report, Elon Musk disclosed that his artificial intelligence venture, xAI, partially utilized models from OpenAI during the development of its Grok chatbot. This disclosure emerges amidst ongoing litigation against OpenAI in federal court, where discussions around governance and AI industry standards are underway.
During a recent testimony at a California federal court hearing, Musk elaborated on his legal battle with OpenAI and its leadership, including CEO Sam Altman and co-founder Greg Brockman. The lawsuit alleges that OpenAI deviated from its initial nonprofit objectives.
When questioned about the use of distillation techniques on OpenAI models by xAI, Musk acknowledged their partial use, highlighting it as a common practice within the AI sector.
Musk was one of the original founders of OpenAI in 2015 alongside Sam Altman and others but left the organization in 2018. Distillation involves training new AI systems using responses from existing models obtained via public APIs or interfaces, serving as learning inputs. In February, Anthropic accused several Chinese AI firms of illicitly harvesting data from its Claude chatbot to train competing products. The White House has also cautioned against large-scale efforts employing proxy accounts and jailbreaks to replicate U.S. AI technologies.
Musk’s testimony reveals that such practices are not limited to international competitors but are adopted by domestic companies as well.
While distillation is not illegal per se, it may contravene certain platform policies or API usage agreements.
xAI was introduced in July 2023 into a competitive landscape dominated by Google, Microsoft, and OpenAI. Earlier that year, Musk had called for a temporary halt on advancing AI development through an open letter co-signed with other tech leaders, pointing to potential risks involved.
Musk’s statements imply xAI might have leveraged technology from his former company to bridge infrastructure gaps.
Neither OpenAI nor xAI responded to Decrypt’s request for comment.