AI Hallucination ResearchFindings by audiencePractitionersInternational / MultilateralLawyersDetail › Finding
Practitioners — Lawyers · Last updated 11 Jun 2026 · Hallucination Register
Share / Print X LinkedIn Email

Finding#1 : EIA screening threshold misattributed to wrong article

RLB Citation ID: RLB-F-INT-UNTC-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-Q001
AI's failure:Misattributed Risk for Lawyers:Liability / PI exposure
What the RLB Specialist Panel found
For Claude Opus 4.7 (web search on)
Question (paraphrased to protect IP)

Which article of the BBNJ Agreement establishes the qualitative screening threshold that triggers an environmental impact assessment obligation for planned high-seas activities?

RLB's analysis

Claude Opus 4.7 displaced the threshold provision by three article numbers, citing Article 30 as the governing obligation and Article 31 as the screening step when the Agreement's authoritative text places both functions at Article 27. The procedural logic the model described is structurally plausible, but its attachment to the wrong article anchors produces a response that appears precise while being positionally incorrect.

AI Head's analysis — what weakness in the AI model caused this

This finding implicates article-level provision mapping in the training data. Claude Opus 4.7 with web search paraphrased the EIA screening test correctly (an activity likely to have more than a minor or transitory effect) but anchored that test to Article 30 of the Agreement and identified Article 31 as the screening step. The verbatim screening-threshold provision sits at Article 27 of Part IV. The error is a clean article-number reassignment that survives substantive review and surfaces only when the citation is checked against the deposited treaty text.

A structured article-by-article provision map for newly-in-force multilateral instruments, applied to corpus ingestion, would address this class of error directly.

For Claude Sonnet 4.6 (web search on)
Question (paraphrased to protect IP)

Which article of the BBNJ Agreement establishes the qualitative screening threshold that triggers an environmental impact assessment obligation for planned high-seas activities?

RLB's analysis

Claude Opus 4.7 displaced the threshold provision by three article numbers, citing Article 30 as the governing obligation and Article 31 as the screening step when the Agreement's authoritative text places both functions at Article 27. The procedural logic the model described is structurally plausible, but its attachment to the wrong article anchors produces a response that appears precise while being positionally incorrect.

AI Head's analysis — what weakness in the AI model caused this

This finding implicates article-level provision mapping in the training data. Claude Opus 4.7 with web search paraphrased the EIA screening test correctly (an activity likely to have more than a minor or transitory effect) but anchored that test to Article 30 of the Agreement and identified Article 31 as the screening step. The verbatim screening-threshold provision sits at Article 27 of Part IV. The error is a clean article-number reassignment that survives substantive review and surfaces only when the citation is checked against the deposited treaty text.

A structured article-by-article provision map for newly-in-force multilateral instruments, applied to corpus ingestion, would address this class of error directly.

Impact for Lawyers in international jurisdictions advising on the BBNJ High Seas Biodiversity Agreement

A lawyer advising a client on whether a planned high-seas activity must undergo an environmental impact assessment under the BBNJ Agreement would, on this AI response, cite the wrong article and could mis-scope the obligation. The screening threshold is set out at Article 27 of Part IV, not Article 30. A legal opinion or transactional disclosure that pins the EIA trigger to Article 30 fails on its face as a treaty citation, and a counterparty or supervisor reviewing the document would identify the error before relying on it.

References — raw findings (per AI model)
This finding also affects
Next finding → Finding#2 : MGR retroactivity default inverted
Cite this finding

Each finding has a stable Citation ID (RLB-F-… for aggregated case-study findings, RLB-H-… for raw per-model hallucinations) — like a DOI, the ID always resolves to the canonical finding even if URLs change.

RLB Citation ID: RLB-F-INT-UNTC-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-Q001
Bluebook / OSCOLA (US + UK legal) Download
RegLeg Specialist Panel, Finding#1 : EIA screening threshold misattributed to wrong article [RLB-F-INT-UNTC-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-Q001], RegLegBrief AI Hallucination Research (June 11, 2026), https://reglegbrief.com/regulators/j1/INT/UNTC/BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023/practitioners/lawyers/finding/INT-UNTC-INT-001-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-v1-001/.
Plain text Download
RegLeg Specialist Panel (2026). "Finding#1 : EIA screening threshold misattributed to wrong article — Practitioners — Lawyers." Citation ID: RLB-F-INT-UNTC-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-Q001. RegLegBrief AI Hallucination Research, published 2026-06-11. https://reglegbrief.com/regulators/j1/INT/UNTC/BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023/practitioners/lawyers/finding/INT-UNTC-INT-001-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-v1-001/
APA 7th edition Download
RegLeg Specialist Panel. (2026). Finding#1 : EIA screening threshold misattributed to wrong article [Hallucination finding RLB-F-INT-UNTC-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-Q001]. RegLegBrief AI Hallucination Research. https://reglegbrief.com/regulators/j1/INT/UNTC/BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023/practitioners/lawyers/finding/INT-UNTC-INT-001-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-v1-001/
BibTeX Download
@misc{reglegbrief_RLB_F_INT_UNTC_BBNJ_HIGH_SEAS_BIODIVERSITY_AGREEMENT_2023_Q001,
  author    = {RegLeg Specialist Panel},
  title     = {Finding#1 : EIA screening threshold misattributed to wrong article},
  year      = {2026},
  publisher = {RegLegBrief AI Hallucination Research},
  note      = {Hallucination finding Citation ID: RLB-F-INT-UNTC-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-Q001},
  url       = {https://reglegbrief.com/regulators/j1/INT/UNTC/BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023/practitioners/lawyers/finding/INT-UNTC-INT-001-BBNJ-HIGH-SEAS-BIODIVERSITY-AGREEMENT-2023-v1-001/}
}
← Back to case study summary Case study detail →

Every finding on this page compares an AI subject's account of the rule against the regulator's verbatim text from the regulator's own portal. Both are linked. Each delta, its root causes, and impact analysis are documented and published with immutable Citation IDs.