What happened
Rio Tinto sued Vale and others in the US District Court for the Southern District of New York. The two companies agreed on a protocol for predictive coding and asked the court to approve it. Rio Tinto would use Relativity Assisted Review through its vendor Precision Discovery. Vale would use Deloitte's Dynamic Review, which classifies documents with the LibLinear library. Each side would check its review against a random control set and disclose estimated recall and precision. Before production, each would sample the documents marked likely non-responsive. On 2 March 2015 Magistrate Judge Andrew J. Peck approved the protocol. He wrote that it is now black letter law that courts permit a producing party to use TAR. He stressed that holding TAR to a higher standard than keywords or manual review deters its use, because motion practice can cost more than TAR saves. He noted that the protocol came from the parties' agreement and was somewhat vague and generic. The court tested neither tool, and the review had not run.
Sources
-
Court document, by authors with no tie to the system.
United States District Court for the Southern District of New York (Magistrate Judge Andrew J. Peck). Rio Tinto PLC v. Vale S.A., 306 F.R.D. 125 (S.D.N.Y. 2015), Opinion and Order of 2 March 2015, with the parties' letter and stipulated protocol attached as an appendix. Caselaw Access Project copy of the Federal Rules Decisions report.
The Court is approving the parties' TAR protocol, but notes that it was the result of the parties' agreement, not Court order.
Details
AI system
- Workflow
-
Rio Tinto and Vale agreed on a predictive coding protocol and asked the court to approve it. Each responding party reviews a random control set, trains the software on seed and training sets and discloses estimated recall and precision. Before production, each party reviews a sample of the documents that the software marks likely non-responsive and discloses the implied recall.
Follow-up
- Conditions for replication
-
The parties disclose to each other the non-privileged documents in their control, seed and training sets. Samples are large enough for a margin of error of plus or minus 2% at the 95% confidence level. The court holds TAR to no higher standard than keywords or manual review.
Entry details
- Sector
-
Legal services / e-discovery.
- Purpose
-
Relativity built Relativity Assisted Review and Deloitte built Dynamic Review to sort documents as likely responsive or not, learning from the coding that lawyers give to samples.
31 more questions not yet checked
model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this record
Peter H. Vartanian. (2026). Federal court approves a TAR protocol agreed by Rio Tinto and Vale (AIBD-0981). Artificial Intelligence Benefits Database, edition of 1 November 2026. https://petervartanian.xyz/aibd/records/AIBD-0981.html (archived at https://doi.org/10.7910/DVN/AADSII, version 7.0)