By Bill Hartzer · Published · How this site is researched
What happened
In a personal injury claim, the defendant insurer ran the claimant's medical records through DigitalOwl, a platform that analyses medical records, and sent the court-appointed orthopedic expert a document titled Body Injury Claim Summary. It contained chronological timelines, highlighted keywords and findings organized by body part. Alongside it went the raw medical documents, bearing coloured highlighting and marginal annotations.
As in the German proceedings above, the expert here was the court's appointee rather than a party's witness. What a party may put in front of such an expert is therefore a question about the integrity of the tribunal's own evidence.
What the court did
The court ordered the expert to disregard the summary document and the annotated raw records, and ordered the insurer to resend the raw medical documents with no markings, marginal notes or additions. It imposed no monetary sanction, treating the issue as novel.
The concern it identified was directed emphasis. Selective highlighting and organization could steer the neutral expert's interpretation and impair the expert's independent judgment and objectivity. Nothing in the summary was alleged to be false. The objection was to what had been foregrounded and what had not.
That is a different failure mode from the ones most of this record documents. There is no fabricated citation, no undisclosed drafting, no unreproducible calculation. A party used a system to decide which parts of a record the expert would see first.
Where it stands
The order is interlocutory and was made on December 9, 2024. Later proceedings were not located as of September 28, 2026.
Read it with Shalom David Biton v. The National Insurance Institute, decided by a Jerusalem labor court ten months later, which allowed a claimant to put ChatGPT-sourced research to a court-appointed expert provided the underlying articles were attached rather than the system's summaries of them. Together the two mark a line: AI-selected source material may reach a neutral expert when the sources come with it; AI-processed and annotated summaries may not.
The obligation it engaged
Clause 1, sequence, is engaged from an unfamiliar direction. The clause bars an expert from asking a system what conclusion the evidence supports. Here a party did something adjacent on the expert's behalf, arranging the evidence so that some findings arrived emphasized. An expert who accepts pre-processed material without knowing how it was selected has let a system shape the sequence whether or not the expert ever used one.
Clause 6, custody and confidentiality, is engaged by the route the records traveled. A claimant's medical file passed through a commercial platform in the course of the litigation, and nothing in the order read suggests the claimant was asked.
Sources
- AI Evidence Database (Damien Charlotin) · September 28, 2026
L. H. S. v. Clal Insurance Co. Ltd. et al.
What could not be confirmed
The decision was read in machine translation from Hebrew; no passage should be quoted without a check against the original. The database captions the matter L. H. S. v. Clal Insurance Co. Ltd.; the document itself reads Plonit, the Hebrew convention for an anonymized female party, and that caption is used here. Whether the expert had already read the summary before the order issued was not determined.
About this entry
This entry was checked against the primary document listed in the sources. The entry for Plonit v. Clal Insurance records conduct and what the tribunal did. It does not characterize a holding and it is not legal advice. Last checked September 28, 2026. More matters like this one are collected under expert witnesses and AI. If something here is out of date, write to [email protected].