Musk v. Altman trial: xAI used ChatGPT to train Grok

Core conclusion

The trial of Elon Musk v. OpenAI will be broadcast live next week, but the newly disclosed evidence has caused an uproar: Musk's xAI company was confirmed to have used the output of OpenAI ChatGPT to train the Grok model.

In other words, Musk was suing OpenAI for "deviating from its non-profit mission" while secretly "distilling" ChatGPT himself.

Key Points

• Source of evidence: Litigation materials disclosed by the court show that xAI engineers clearly recorded the use of ChatGPT API output to fine-tune Grok in internal documents.
• Time involved: 2025 from the early days of xAI’s establishment to the release of Grok 1.0
• Core Change: The core argument of Musk v. OpenAI is that "OpenAI makes the GPT model closed source, which violates the original intention of open source non-profit." However, xAI's reliance on ChatGPT directly proves that OpenAI's model itself has sufficient public value.
• Industry Impact: This case has become a landmark case on "whether model distillation is infringement" in the AI era

Background and trigger events

Musk OpenAI is one of the most dramatic legal showdowns in the AI industry. Elon Musk was a co-founder of OpenAI, co-founding the company in 2015 before leaving in 2018 after a failed battle for control. In 2024, Musk filed a lawsuit accusing OpenAI of abandoning its non-profit mission, being overly tied to Microsoft, and asking the court to prevent OpenAI from transforming into a for-profit company.

This case has been dominated by written materials before, but next week it will be broadcast live via YouTube for the first time (11AM-5PM ET), which is called the "AI Trial of the Century" by the media.

The latest evidence disclosed comes from internal communications records required by the court to be submitted by both parties. xAI’s internal Slack channel and engineer emails show that in the early stages of 2025, the xAI team believed that “distilling from ChatGPT is the fastest way to get started.”

Key Impact

Adaptation suggestions

For AI content creators

• API usage compliance first: The terms of service for most AI APIs (OpenAI, Anthropic, Google) explicitly prohibit using the output to train competing models. xAI’s actions remind us: Regulation is not a slogan, it is a contract.
• Local model priority strategy: If you need a lot of fine-tuning or distillation, give priority to locally deployed open source models (such as DeepSeek V4, Llama 4, Qwen) to avoid legal risks.
• Pay attention to the trend of the case: The judgment in this case will directly affect the API usage specifications of the AI ​​Agent industry, and it is recommended to continue to follow it.

Shift in focus

If you are using DeepSeek V4 to run Claude Code to reduce content production costs, this news gives you an additional reason: Open source model + local deployment = zero compliance risk.

Related extended information

• YouTube: （）
• HN

Tool entry

Tool names that appear naturally in the text: OpenAI, ChatGPT, DeepSeek, Claude, Grok, xAI. The platform side will match the maintained tools library.

Internal link guidance

• Want to replace high-priced APIs with open source solutions? Watch: DeepSeek V4 vs Claude Code: 90% Cheaper, Same Quality
• Compliance case for AI entrepreneurship: Claude Code 48 hours to start a business: one person + US$29 monthly fee, monthly income in 3 months $9,000