OpenAI Seeks to Disqualify Expert Etzioni Over Unverified AI Charts
Lawyers for OpenAI and Microsoft argue that Professor Oren Etzioni’s expert report relies on hallucinated data, contradicting his own advice to verify AI outputs.
Key points
- OpenAI and Microsoft filed a motion to exclude Professor Oren Etzioni’s expert testimony in the MDL lawsuit.
- The defendants argue Etzioni’s charts were generated by AI and contained fabricated statistics and quotes.
- Etzioni previously warned that failing to verify AI results is equivalent to gambling with the technology.
Lawyers representing OpenAI and Microsoft have moved to disqualify Professor Oren Etzioni as an expert witness in the multi-district litigation concerning AI training data. According to filings reported by Techdirt, the defendants argue that Etzioni’s opinions are unreliable because they depend on AI-generated charts that have since been withdrawn.
The motion filed last month asserts that Etzioni’s figures included hallucinated data derived from cited academic papers. Techdirt reports that the legal teams for the tech giants contend these fabricated statistics undermine the validity of his testimony regarding the importance of news content in large language model training.
Context of the Disqualification
This attempt to exclude Etzioni is part of a broader effort by OpenAI and Microsoft to disqualify several plaintiff experts in the ongoing lawsuit. The case involves over 2,100 filings, making it difficult to track every procedural move. Techdirt notes that while challenging expert credentials is common, the specific claim against Etzioni is notable due to his public stance on AI reliability.
Etzioni, a University of Washington professor and founding CEO of the Allen Institute for A.I., has long warned about AI risks. He previously stated that AI models can fabricate quotes and invent statistics with false confidence. His own nonprofit, TrueMedia.org, was founded in January to combat the spread of disinformation and deepfakes.
Who Is Affected
The outcome of this motion could influence how courts evaluate expert testimony in AI-related litigation. Etzioni’s credibility is central to the plaintiffs’ argument about the value of news content in training models. If excluded, it may set a precedent for scrutinizing AI-assisted research in legal proceedings.
Techdirt highlights the irony that Etzioni, who warned against relying on unverified AI output, faced disqualification for failing to verify his own charts. His previous warnings emphasized that skipping verification steps turns reading into gambling with AI.
What should you do?
Legal teams and researchers using AI tools for report generation should implement rigorous verification protocols. Professionals must cross-check all AI-generated statistics, quotes, and figures against primary sources before including them in official documents or expert reports.
Background: Large language models
Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.
What to do and how to stay safe: OpenAI
- Verify all AI-generated statistics and quotes against original academic papers or primary sources.
- Implement a manual review process for any data visualizations created with AI assistance.
- Maintain documentation of the verification steps taken to ensure transparency in expert reports.
- Avoid including AI-generated content in legal or professional submissions without independent confirmation.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Why is OpenAI trying to disqualify Oren Etzioni?
OpenAI and Microsoft argue that Etzioni’s expert report relies on AI-generated charts containing hallucinated figures and fabricated statistics.
What did Etzioni say about verifying AI output?
Etzioni warned that users must always verify important information themselves because AI models can fabricate quotes and invent statistics with false confidence.
How many filings are in the MDL lawsuit?
According to Techdirt, there are currently over 2,100 filings in the multi-district litigation against OpenAI and Microsoft.
