Lead story
New York court treats a pro se litigant's AI prompts as work product
A New York trial court quashed subpoenas seeking a self-represented defendant's prompts, uploads, and outputs from OpenAI, finding the requested material protected as litigation preparation.
- What happened
- A New York trial court quashed subpoenas seeking a self-represented defendant's prompts, uploads, and outputs from OpenAI, finding the requested material protected as litigation preparation.
- Why it matters
- AI prompts and outputs can reveal litigation strategy even when no lawyer created them. Discovery plans should identify whether AI materials are relevant, who created them, whether they were prepared for litigation, what the provider retained, and whether disclosure to the service affects protection under the controlling jurisdiction.
- Watch
- The analysis was published October 27, but the ruling was issued June 4, 2026. The decision applies New York law to a civil case and does not establish a uniform rule; other courts have reached different results, particularly on waiver and in criminal matters.
Sourcesedrm.netscholar.google.com
Briefing item
Court requires email-by-email detail in a privilege log
A federal court ordered a party to revise a privilege log that grouped long email strings under single entries, following the majority approach that each withheld email must be logged at least once.
- What happened
- A federal court ordered a party to revise a privilege log that grouped long email strings under single entries, following the majority approach that each withheld email must be logged at least once.
- Why it matters
- Thread-level review and logging can reduce cost, but a top-email-only entry may conceal who received an earlier message or whether every component is protected. Teams should settle the logging unit, treatment of duplicates and attachments, and any categorical protocol during discovery planning rather than after a challenge.
- Watch
- The analysis was published October 27; the cited order was issued June 23, 2026. The Sixth Circuit has not resolved the email-string question, and courts remain divided, so the required level of detail depends on the jurisdiction and any agreed protocol.
Sourcesedrm.netscholar.google.com
Briefing item
Patient-data incident puts third-party deletion proof under scrutiny
Johns Hopkins Bayview notified about 3,182 patients after an employee uploaded a spreadsheet to an unapproved public AI tool, while public reporting left unresolved whether the tool's operator completed the requested deletion.
- What happened
- Johns Hopkins Bayview notified about 3,182 patients after an employee uploaded a spreadsheet to an unapproved public AI tool, while public reporting left unresolved whether the tool's operator completed the requested deletion.
- Why it matters
- A user's deleted chat is not proof that all provider copies, review queues, logs, or backups are gone. Incident and legal teams should preserve evidence of what was sent, document the deletion request and response, determine whether legal or safety retention applies, and retain the source record and incident file when preservation duties may exist.
- Watch
- The October 27 report synthesizes hospital statements, news accounts, and federal rules. The AI provider was not identified, public reporting did not establish that it confirmed deletion, and the HHS breach portal may lag submitted reports. The incident details remain subject to further disclosure.
Sourcescomplexdiscovery.combeckershospitalreview.comecfr.govecfr.gov
Briefing item
ProSearch certifies its AI management system under ISO 42001
ProSearch announced ISO/IEC 42001 certification for the AI management system supporting its eDiscovery operations, including controls for client-data handling and AI development, testing, and approval.
- What happened
- ProSearch announced ISO/IEC 42001 certification for the AI management system supporting its eDiscovery operations, including controls for client-data handling and AI development, testing, and approval.
- Why it matters
- Legal teams evaluating AI-enabled discovery providers can use the certification scope, statement of applicability, audit cycle, risk process, and exceptions as diligence inputs. Matter-level defensibility still requires testing outputs, documenting workflows, and confirming that contracted services and data flows fall within the certified system.
- Watch
- This is a commercial announcement published October 26 and covered October 27 by a sponsored publication. Certification supports claims about a management system, not the legal defensibility, security, or performance of every use case; the underlying audit materials were not publicly reviewed.
SourceseDiscovery Todayprosearch.comiso.org
Briefing item
Casepoint adds a source-linked investigation agent
Casepoint announced an AI investigation workflow that narrows document populations, extracts facts with citations, links evidence to themes, and supports timelines and findings reports in its eDiscovery platform.
- What happened
- Casepoint announced an AI investigation workflow that narrows document populations, extracts facts with citations, links evidence to themes, and supports timelines and findings reports in its eDiscovery platform.
- Why it matters
- Source citations can make AI-assisted findings easier to verify, but they do not establish completeness. Investigation teams should test retrieval across each data type, sample material the system ranks low, preserve prompts and workflow history, and require reviewers to confirm both inculpatory and exculpatory evidence before relying on a report.
- Watch
- This is a commercial product announcement reproduced by a sponsored publication. General availability is planned for November, and claims about scale, speed, coverage, and performance were not supported by an independent benchmark in the cited materials.
SourceseDiscovery TodayCasepoint
Educational only. Not legal advice. Review sources and matter-specific obligations before relying on any workflow decision.