Evgeny Yaroshevsky v. Karlin
U.S. District Court, Eastern District of New York · E.D.N.Y. · New York bar guidance
Conduct
Pro se plaintiff admitted using Gemini for citations, filed at least 12 fabricated cases plus 13 fake quotations.
Consequence
Magistrate Judge recommended an adverse costs order against the pro se plaintiff.
Lesson
Adverse costs orders, not just warnings, are now in play for pro se litigants who admit AI use without verification in E.D.N.Y.
Verified May 14, 2026
- Citation
- Yaroshevsky v. Karlin, No. 1:25-cv-02747 (E.D.N.Y. Feb. 23, 2026) (Eichenholtz, M.J.)
- Decided
- February 23, 2026
Summary
Pro se plaintiff Evgeny Yaroshevsky admitted using Google Gemini to obtain case citations and failed to verify them before filing. Defendants identified at least seven non-existent cases in the complaint and five in his opposition to the motion to dismiss, plus thirteen instances of quoted language attributed to cases that does not appear in the cited opinions. Magistrate Judge Seth D. Eichenholtz issued a Report and Recommendation finding the citations fabricated and recommending an adverse costs order against the plaintiff.
- AI tool:
- Google Gemini
What sanction did the court impose?
Magistrate Judge recommended an adverse costs order against the plaintiff. No bar referral (pro se litigant).
Why does Evgeny Yaroshevsky v. Karlin matter for law firms using AI?
Yaroshevsky is an important inflection point for AI-hallucination sanctions practice in the Second Circuit’s district courts. Where S.D.N.Y. has settled into a warning-first pattern for first-occurrence pro se hallucinations, E.D.N.Y. through Magistrate Judge Eichenholtz’s R&R has shown willingness to recommend adverse costs orders against pro se litigants who (a) admit AI use on the record and (b) submit a high volume of hallucinations (here, at least twelve fabricated cases plus thirteen false quotations). The combination of admission and volume distinguishes the disposition from the warning-only pattern.
For firms with E.D.N.Y. matters opposed by pro se filers, the strategic takeaway is that adverse costs requests are now part of the sanctions toolkit, not just motions to strike. The procedural posture matters: the R&R mechanism gives the magistrate judge room to develop the record on the hallucinations before the District Judge rules, which gives opposing counsel a meaningful window to brief the sanctions question before adoption. Firms should treat first detection of AI-fabricated citations as the trigger for a sanctions-position memo, not just a strike motion.
Implications for your firm
Operational steps a firm reading this case may wish to consider documenting. Strategic and rule-application calls belong to your firm's attorneys.
- On-the-record admission of AI use without verification is escalating the sanctions calculus from warning to adverse costs even for pro se litigants. The warning-only floor in the S.D.N.Y. cases (Tantaros, Anonymous v. NYC DOE) does not appear to control in E.D.N.Y. when admission is on the record.
- When opposing a pro se filer who admits AI use in correspondence or on the record, an adverse costs request is now a viable ask in E.D.N.Y., not just a move to strike.
- Identifying both fabricated case citations and fake quotations in the same filing strengthens the sanctions request; the dual hallucination pattern is what drove the Yaroshevsky R&R.
Sources
Primary sources
Further reading
- The on-file primary source is Magistrate Judge Eichenholtz's Report and Recommendation (filed 2026-02-23). Whether the District Judge subsequently adopted the R&R, and on what date, is not yet verified against an adoption order; the Charlotin tracker notes the R&R 'was adopted later on' but does not link it.