{"schema_version":"0.8.0","id":"LR-2026-0102","record_version":3,"title":"Connecticut pro se plaintiff sanctioned for prompt injection against court AI tool","verification":{"human_verified":false,"method":"automated_gate","auto_verified":true,"score":46.2,"score_breakdown":{"source_authority":12.1,"independent_corroboration":7.0,"implicated_party_confirmed":0.0,"citation_integrity":14.0,"specificity":0.0,"completeness":5.76,"unhedged_language":5.0},"checks":{"schema_valid":true,"every_field_cited":true,"quotes_verbatim":true,"source_fetched":true,"distinct_incident":true,"dates_plausible":true,"sources_agree":true},"gate_version":"gate-2026-09-29.5","cross_check_model":null,"cross_check_agreement":null,"duplicate_of":null,"verified_by":null,"verified_at":"2026-09-29","extraction_run_id":null},"system":{"model_name":null,"model_version":null,"developer":"unknown","harness":null,"tools_granted":[],"permissions_scope":null,"autonomy_level":"unknown","guardrails_present":[],"deployment_context":"unknown","tier":null,"lifecycle_stage":null,"integration_type":null,"intended_use":null,"safeguard_state":null,"production_intended":null},"event":{"date_occurred":null,"date_disclosed":"2026-09-29","incident_class":"security_compromise","failure_class":null,"agent_involved":false,"secondary_failure_classes":[],"trigger":"prompt_injection","duration":null,"detection_method":"unknown","summary":"A Connecticut pro se plaintiff hid instructions aimed at a court AI tool inside court filings, attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions. The court caught the attempt, held that the prompt injection attack offended the integrity of the proceedings, and revoked the plaintiff's electronic filing privileges.","tags":[],"date_detected":null,"near_miss":true,"root_cause":"adversarial_input","root_cause_detail":null,"failure_class_detail":null,"responsible_party":null,"adversarial_actor":"external_attacker","input_provenance":"court filings","time_to_contain":null,"operations_paused":null,"remediation":"The court revoked the plaintiff’s electronic filing privileges.","precursor_events":null},"impact":{"direct_loss_usd":null,"remediation_hours":null,"records_exposed":null,"third_parties_affected_count":null,"severity":"minor","severity_because":"minor: The incident was contained to a single court proceeding with routine cleanup—the plaintiff's filing privileges were revoked—and no third party was touched.","harm_class":[],"affected_description":null,"protected_class_involved":null,"data_categories_exposed":[],"loss_range_usd":null,"reversibility":null,"downstream_exposure_count":null},"evidence":{"sources":[{"url":"https://www.jdsupra.com/legalnews/ai-quarterly-a-review-of-ai-law-policy-3253293/","publisher":"news search (Bing News index)","date":"2026-09-29","type":"press","title":"AI Quarterly | A Review of AI Law, Policy & Practice | 2026 Q3 | JD Supra","archived_url":null,"attachment":"first_report"}],"per_field_confidence":{"system.developer":"D","system.autonomy_level":"D","system.deployment_context":"D","event.near_miss":"D","event.root_cause":"C","event.adversarial_actor":"C","event.input_provenance":"C","event.remediation":"C","event.agent_involved":"D","event.incident_class":"C","event.trigger":"C","event.summary":"D","event.date_disclosed":"A","event.detection_method":"D","impact.severity":"D","context.industry":"D","context.denominator_available":"D","context.country":"D","impact.severity_because":"D"},"field_citations":{"event.incident_class":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.trigger":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.adversarial_actor":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.input_provenance":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.agent_involved":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.root_cause":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.near_miss":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.remediation":[{"source_index":0,"quote":"The court held that the plaintiff’s attempted use of an AI “prompt injection” attack offended the integrity of the proceedings, and revoked the plaintiff’s electronic filing privileges."}],"context.industry":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"context.country":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"event.summary":[{"source_index":0,"quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings."}],"impact.severity":[{"source_index":0,"quote":"The court held that the plaintiff’s attempted use of an AI “prompt injection” attack offended the integrity of the proceedings, and revoked the plaintiff’s electronic filing privileges."}],"impact.severity_because":[{"source_index":0,"quote":"The court held that the plaintiff’s attempted use of an AI “prompt injection” attack offended the integrity of the proceedings, and revoked the plaintiff’s electronic filing privileges."}],"event.date_disclosed":[{"source_index":0,"quote":"published: 2026-09-29"}]},"notes_on_gaps":"No source states the model name, version, developer, harness, tools granted, permissions scope, autonomy level, guardrails present, deployment context, lifecycle stage, integration type, intended use, safeguard state, production intent, date occurred, date disclosed, failure class, secondary failure classes, duration, detection method, tags, date detected, root cause detail, responsible party, time to contain, operations paused, precursor events, incident class detail, direct loss USD, remediation hours, records exposed, third parties affected count, harm class, affected description, protected class involved, data categories exposed, loss range USD, reversibility, downstream exposure count, org size band, denominator, systemic exposure, shared model id, upstream incident id, incident evidence, regulatory, or coverage. The court AI tool is described only as an AI tool the court 'believed was relying on' for decisions; no vendor or model is named. The plaintiff is a pro se individual, not an organisation, so org size band and denominator are not applicable. The incident is a near miss: the injection attempt was caught and sanctioned before any loss materialised. The severity is minor because the incident was contained to a single court proceeding with routine cleanup (revocation of filing privileges) and no third party was touched. The adversarial actor is the plaintiff, an external attacker relative to the court. The root cause is adversarial_input because untrusted content (court filings) carried the instruction. The trigger is prompt_injection because instructions embedded in content the system read set it off. The incident class is security_compromise because an injected instruction caused the system to act. No agent was involved; the AI tool was a decision-support system, not an autonomous agent with tools. The detection method is unknown because no source states who first noticed the injection attempt; the court's sanction implies detection but does not name the noticer or means. The date occurred and date disclosed are not stated; the source is dated 2026-09-29 but that is the publication date, not necessarily the incident date. The failure class is null because no autonomous agent was involved. The story answers are drawn from the source's stated answers; the 'caught' step names the court as the noticer, but the source does not specify the means of detection, so the detection method remains unknown. system.tier: not set; no operator named and deployment_context is not a lab's. [aggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29] system.developer: no source names the developer. system.autonomy_level: no source states the arrangement. system.deployment_context: no source states where it ran. event.near_miss: classified from described behaviour. event.agent_involved: classified from described behaviour. event.summary: classified from described behaviour. event.detection_method: no source says who first noticed. context.industry: classified from described behaviour. context.denominator_available: no denominator is available for this incident. context.country: classified from described behaviour. [corrected by Srini Iyer, approved 2026-10-18 in the corrections review: severity reason printed the whole rubric]"},"context":{"industry":"Public Administration","org_size_band":null,"country":"US","denominator_available":false,"denominator":null,"related_incident_ids":[],"systemic_exposure":null,"shared_model_id":null,"upstream_incident_id":null,"incident_evidence":null,"regulatory":null,"coverage":null},"narrative":{"steps":[{"step":"doing","text":"A Connecticut court was using an AI tool to help reach decisions, and a pro se plaintiff filed documents with it. The plaintiff hid instructions aimed at that tool inside court filings.","quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings.","source_index":0,"grade":"C"},{"step":"went_wrong","text":"The plaintiff tried to manipulate the court's AI tool by embedding instructions in filings, a prompt injection attack that offended the integrity of the proceedings.","quote":"The court held that the plaintiff’s attempted use of an AI “prompt injection” attack offended the integrity of the proceedings, and revoked the plaintiff’s electronic filing privileges.","source_index":0,"grade":"C"},{"step":"reached","text":"The attempt targeted the court's AI decision tool through the court's electronic filing system; the court caught it and revoked the plaintiff's electronic filing privileges.","quote":"The court held that the plaintiff’s attempted use of an AI “prompt injection” attack offended the integrity of the proceedings, and revoked the plaintiff’s electronic filing privileges.","source_index":0,"grade":"C"},{"step":"caught","text":"The court caught the plaintiff's attempt and sanctioned him, holding the prompt injection attack offended the integrity of the proceedings.","quote":"A Connecticut pro se plaintiff has been caught—and sanctioned for—attempting to manipulate a court he believed was relying on an artificial intelligence tool to reach decisions by hiding instructions directed to the tool in court filings.","source_index":0,"grade":"C"},{"step":"done","text":"The court sanctioned the plaintiff and revoked his electronic filing privileges. The episode was framed as a cautionary tale about AI security and judicial concerns over AI in legal proceedings.","quote":"The court held that the plaintiff’s attempted use of an AI “prompt injection” attack offended the integrity of the proceedings, and revoked the plaintiff’s electronic filing privileges.","source_index":0,"grade":"C"}],"model":"deepseek-chat","prompt_version":"aggregate-2026-09-27.1","extracted_at":"2026-09-29","source_urls":["https://www.jdsupra.com/legalnews/ai-quarterly-a-review-of-ai-law-policy-3253293/"]}}