For the first time, a federal court filing has thrust generative AI (GenAI) into the mainstream conversation for legal disclosure. In EEOC v. Tesla, Inc., the parties submitted a proposed discovery protocol that referenced GenAI as a mutually agreed approach for responsiveness review.

While this is likely not the first time GenAI has been deployed for document review, the Tesla decision is one of the earliest, if not the first, court-recognised example of AI document review available on the public record worldwide. This development may mark the tipping point for widespread court acceptance of GenAI in legal disclosure protocols, building upon established global jurisprudence regarding technology assisted review (TAR) and signaling a new phase of eDiscovery innovation.

This article considers what the Tesla decision tells us about where the legal market is heading with GenAI, and how lessons from the established use of TAR in the US, UK, and Australia lay the groundwork for broader acceptance.

Tesla: A Case Study in GenAI Disclosure

In Tesla, both parties acknowledged the availability of advanced analytics, including TAR and GenAI, as part of their document review process. Importantly, the protocol includes explicit requirements for statistical validation and ongoing transparency. 

The relevant section states:

"The parties also recognise the availability of a variety of search tools and methodologies, including but not limited to Technology Assisted Review (TAR) and Gen AI tools. Tesla has notified plaintiff’s counsel that it may use TAR and/or GEN AI tools to further analyse documents for relevance after search terms are used to narrow the starting document universe to exclude documents not likely to be relevant. If the producing party intends to use TAR, GEN AI, or similar advanced analytics as a substitute for attorney responsiveness review, the parties agree to meet and confer in good faith to attempt to reach agreement about the technology and process that a producing party proposes to use to identify responsive ESI and a statistically sound methodology to determine the recall rate and other measures of the effectiveness of the tool and processes in identifying responsive documents. The producing party shall make disclosures regarding its tools and processes necessary to make the meet and confers meaningful and for the requesting party to negotiate on an informed basis."

The protocol highlights several themes: 

• The growing expectation that GenAI may play a central role in responsiveness review
• The importance of openness in the choice and deployment of technology
• An insistence on evidence-driven validation

Statistical evaluation— measuring recall, precision and F-score— is fundamental, ensuring legal teams keep validation and defensibility top of mind throughout the process. 

How GenAI Review Works and How Results Are Validated

GenAI for document review relies on large language models (LLMs) to evaluate the content and relevance of documents. The value is clear: the ability to perform high-volume review faster, at a lower cost, and often with higher accuracy than human reviewers.

Platforms such as Relativity’s aiR for Review, eDiscovery AI, and Reveal supply systems that use natural-language prompts to analyze and code documents for responsiveness to specific issue criteria. Some of these systems also generate document summaries and explanations of the coding decisions applied to the document. 

Statistical validation of the review process is essential for accuracy. The core standard for validating GenAI outputs remains statistical testing:

• Recall: measures how many relevant documents are retrieved out of all relevant documents in the dataset.
• Precision: measures how many of the retrieved documents are actually relevant. 

Legal teams have relied on these metrics for over 15 years to demonstrate the defensibility of AI-driven reviews (TAR or CAL). This tried-and-tested methodology underpins GenAI’s claim to legitimacy. 

Tesla as a Potential Paradigm Shift—Lessons from TAR Jurisprudence

Tesla may represent the beginning of a paradigm shift as GenAI moves beyond its early use for drafting and summarization to become an accepted tool for issue coding in document-heavy matters. This evolution builds directly on earlier court decisions that established TAR as defensible, cost-effective, and accurate.

Some examples:

• In the US, Da Silva Moore v. Publicis Groupe and Rio Tinto v. Vale solidified the principle that technology-assisted workflows, when validated, are at least as reliable as traditional review. 
• In the UK, Pyrrho Investments Ltd v. MWB Property Ltd signalled judicial support for proportional, technology-enabled review, noting cost control and efficiency as primary drivers. Following this decision, disclosure protocols in the Civil Procedure Rules (Practice Direction 57D) were updated to require parties to adopt TAR by default in cases involving large volumes of disclosure, or be prepared to explain why they chose not to.  
• Similarly, McConnell Dowell Constructors v. Santam Ltd out of Australia echoed these themes, approving predictive coding precisely because of its cost-effectiveness in complex disputes. 

A commonality between these decisions has been the tendency to define TAR widely, rather than referring to the specific algorithms, software, or vendors powering TAR. For example, in Pyrrho, Master Matthews notes: 

"the term 'predictive coding' is used interchangeably with 'technology assisted review,' 'computer assisted review,' or 'assisted review.' It means that the review of the documents concerned is being undertaken by proprietary computer software rather than human beings."

Generously interpreted, these vague definitions can be seen as a forward-thinking recognition that technology would evolve rapidly. Their main directive remains clear: use technology to increase efficiency while maintaining or even improving accuracy. 

The reasons courts accepted technology assisted review (TAR)— controlling costs, improving efficiency, and delivering accurate results— apply equally to GenAI, as long as legal teams exercise similar oversight and validation.

Three Open Questions for GenAI Adoption in Review

1. Transparency: The Tesla decision calls for "The producing party [to] make disclosures regarding its tools and processes necessary to make the meet and confers meaningful and for the requesting party to negotiate on an informed basis." How much detail do parties need to share about GenAI workflows, underlying models, vendors, and processes? What strikes the right balance between protecting proprietary information and enabling informed negotiation? 
    
2. Agreement on LLMs Used: While one can easily envisage parties specifying the use of LLMs provided by OpenAI or Meta, is there any real need to specify the actual LLM powering GenAI review, given established TAR protocols do not require disclosure of underlying algorithms? Can statistical validation alone suffice?
    
3. Prompts and Review Protocols: Will parties (or courts) begin to require disclosure of prompts and configuration settings used to guide GenAI review, despite the fact that detailed TAR review protocols are rarely shared?

Three Practical Steps for Legal Counsel Using GenAI for Document Review

While the exact requirements for the use of GenAI will likely differ from jurisdiction to jurisdiction, we expect most, if not all, considerations will involve the following best practices: 

1. Engage early and openly with opposing counsel on the technologies you intend to use (without necessarily being too granular), ensuring meet-and-confer discussions are meaningful and well-informed.
    
2. Insist on— and document— statistical validation of GenAI review outcomes, using sample testing and recall and precision metrics to build defensibility.
    
3. Keep clients informed about both the cost reductions GenAI can achieve and the need for careful oversight throughout the review process.

Conclusion

Tesla may be the first court-recognized reference to GenAI in disclosure protocols, but it is almost certainly not the last. The integration of GenAI into document review builds directly on the legal standards that established TAR as a global norm. 

If legal teams maintain transparency, focus on validation, and engage openly with the courts and their clients, GenAI is poised to become an accepted— and eventually expected— part of modern litigation disclosure.

TransPerfect Legal’s consulting teams are working closely with clients to deploy GenAI in an effective and defensible manner across ongoing matters. To learn more, please reach out to our team.