Professor Withdraws Hallucinated AI Charts from Lawsuit
The retraction follows the expert's failure to verify AI-generated figures used as evidence against OpenAI and Microsoft.
Updated on Oct. 21, 2026 in Artificial Intelligence
Live Poll
Do you personally verify information obtained from AI tools before relying on it?
Professor Oren Etzioni has withdrawn AI-generated charts from his expert report in a lawsuit against OpenAI and Microsoft after admitting the figures contained hallucinated data. Etzioni, who founded TrueMedia.org to combat disinformation, failed to verify output generated by Anthropic's Claude before including it as evidence.
Why it matters
The incident highlights the risks of using generative AI for empirical research in high-stakes legal proceedings, demonstrating how unverified model output can undermine expert credibility. It serves as a reminder that LLMs may synthesize plausible but unsupported data, creating potential for academic and legal error.
The expert report contained figures generated by Anthropic's Claude that lacked support in the underlying academic literature cited by the expert. Etzioni confirmed that the numbers in the report were not cross-referenced against original source data.
The players
Oren Etzioni
Professor and founder of TrueMedia.org, a nonprofit focused on identifying and mitigating AI-generated disinformation.
Anthropic
Developer of the Claude series of large language models, which focus on safety and performance in enterprise and research contexts.
OpenAI
Developer of GPT-series models and a primary defendant in the current litigation regarding AI training data.
Microsoft
A major investor in OpenAI and partner in deploying LLM technology, also named as a defendant in the legal proceedings.
The details
Etzioni employed Anthropic's Claude, a large language model designed to process and synthesize text, to produce data figures intended to support his arguments regarding news content in LLM training. Because LLMs are probabilistic systems—generating content based on statistical patterns rather than retrieving verified facts—the tool hallucinated figures that did not exist in the source material. Etzioni retracted these charts following his deposition, during which the discrepancies were identified.
Timeline
2019: Etzioni warned that AI would accelerate online disinformation spread.
January 2026: Etzioni founded the nonprofit TrueMedia.org.
September 2026: OpenAI and Microsoft filed a motion to exclude expert testimony.
The Tech Race
The withdrawal of this testimony is a significant setback in the broader legal battle between plaintiffs and AI developers over LLM training practices. It reflects the ongoing challenge of maintaining scientific rigor in legal filings while using AI tools in the expert research process.
The incident serves as a warning for researchers and legal professionals who rely on LLMs for data synthesis or evidence preparation. It underscores the necessity of manually verifying every AI-generated claim against primary source material to prevent the introduction of erroneous information into formal workflows.
The takeaway
This case illustrates the critical importance of human verification when applying LLM outputs to formal evidence or professional reports. Researchers should track the outcome of the defense motion to exclude experts, as it will likely influence how courts evaluate AI-assisted evidence in future copyright and training data litigation.
Further reading
Explore more on the evolving standards for synthetic content and validation in Artificial Intelligence.
Source note: This article includes information reported by Techdirt.
Live Poll
Do you personally verify information obtained from AI tools before relying on it?
