AI on BBNJ High Seas Biodiversity Agreement for Professional Engineers in international jurisdictions

Executive Summary

The BBNJ Agreement — the UN treaty on the conservation and sustainable use of marine biological diversity of areas beyond national jurisdiction — establishes binding obligations for environmental impact assessment of high-seas activities, including specific screening thresholds that determine when a full EIA is required. Professional Engineers working on ocean infrastructure, deep-sea resource projects, submarine cable installations, or marine research programmes in international waters have a direct compliance interest in these thresholds and the article references that govern them.

In our testing, AI tools produced an incorrect answer on the EIA screening provision: the trigger standard was subtly but materially restated, and the article citation was wrong. That single error is enough to cause an engineer to mis-scope a compliance obligation before a project leaves the design phase.

How AI gets this regulation wrong

The failure pattern on this regulation is one where AI tools answered with apparent confidence, then revised or qualified the response when pressed — a signal that the initial answer was not well-grounded. In the finding documented here, the error took the form of a subtly narrowed legal standard and a wrong article number: small-looking changes that would pass unnoticed without independent source verification, but that carry significant compliance consequences in practice.

What that means for your practice

For Professional Engineers, the dominant risk from AI errors on the BBNJ Agreement falls squarely in the regulatory enforcement category: getting the EIA screening threshold wrong means mis-advising on whether a proposed activity needs assessment at all, with direct consequences for project authorisation and state-party compliance obligations. The table below maps that risk to the specific finding documented in this cell.

When this affects Professional Engineers

Professional Engineers encounter the BBNJ Agreement most directly when scoping activities that occur in or affect areas beyond national jurisdiction — the high seas and the deep seabed outside any state's exclusive economic zone. This includes submarine telecommunications cable routing and repair, offshore energy infrastructure with a high-seas footprint, scientific research vessel operations, deep-sea resource surveys, and environmental baseline studies supporting marine spatial planning. In all of these contexts, the EIA screening question arises early: does this activity cross the threshold that triggers a formal impact assessment under the Agreement?

When an engineer or their team uses an AI tool to answer that question — whether drafting a technical opinion, preparing project documentation for a flag state, or advising a client on whether EIA obligations apply — an incorrect threshold standard or a wrong article citation can lead to a professional opinion that is factually wrong. If a project proceeds on the basis that it falls below the EIA trigger when it does not, the sponsoring state may face compliance exposure under the Agreement, and the engineer's advice may be shown to have relied on a misstated rule.

The BBNJ Agreement is recent (opened for signature in 2023 and not yet universally in force at the time of this research), which increases the risk that AI tools have incomplete or inaccurate training data on its specific provisions. Professional Engineers working at the frontier of international environmental law for ocean-sector projects cannot treat AI output on this treaty as reliable without primary-source verification.

The findings at a glance

The table below summarises the finding documented in this cell, covering the question area, the nature of the AI's error, and the risk category it creates for Professional Engineers in international jurisdictions.

Aggregate impact

The single finding in this cell targets what is arguably the most consequential practical question the BBNJ Agreement poses for engineers: at what point does a planned activity legally require an environmental impact assessment? The Agreement's text sets the threshold at activities that "may have more than a minor or transitory effect on the marine environment, or the effects of which are unknown or poorly understood." This is an explicitly precautionary formulation — the word "may" and the inclusion of activities with uncertain effects both push the trigger lower than a standard probability-of-harm test would.

When AI tools restate that threshold using "likely to have" instead of "may have," the practical effect is to narrow the class of activities that appear to require assessment, potentially excluding projects that the Agreement's drafters intended to capture.

The article mis-citation compounds that error. An engineer who receives both a subtly wrong standard and a wrong article reference has two independent failure points to catch: the substantive rule is wrong, and the source pointer would not lead to the correct provision even if they went to verify. Verification via the cited article would appear to confirm the AI's answer — a trap that is particularly dangerous for practitioners who are unfamiliar with the Agreement's internal structure and are relying on AI to orient them in a new regulatory text.

For Professional Engineers working in international project teams or advising clients across multiple jurisdictions, the systemic implication is clear: the BBNJ Agreement is exactly the kind of recent, technically specific, and less-widely-cited treaty where AI training data is most likely to be incomplete, and where confident-sounding but wrong answers are most likely to pass unchallenged by non-specialist readers.

What your team should do

The default position for Professional Engineers consulting AI tools on the BBNJ Agreement's EIA provisions should be to treat any threshold standard or article citation as unverified until checked against the treaty text directly. The agreement is available through the United Nations Treaty Collection at treaties.un.org; the EIA chapter is self-contained and the screening article is accessible in the official text. Given that the specific error documented here — a narrowed threshold and a wrong article number — would be invisible without side-by-side comparison with the source, primary-source verification is not optional for opinions that will inform project decisions.

For scoping work, engineers should pay particular attention to the precautionary language in the treaty's EIA screening provision. The "may have" formulation combined with the explicit inclusion of activities with unknown or poorly understood effects means the screening net is wider than a conventional probability-based standard. AI tools that restate this as "likely to have" are applying a different and higher bar, which will systematically undercount activities that fall within the Agreement's intended scope. Any project screening that relies on this threshold should note the specific treaty text, not a paraphrase from an AI tool.

AI tools remain useful for orientation tasks on the BBNJ Agreement — getting an overview of the treaty's structure, understanding how the EIA chapter fits within the broader framework, or identifying which state parties have obligations under a given provision. The risk is concentrated at the point where AI output is used to state specific legal standards or cite specific articles in professional documentation. At that point, the text of the Agreement is the only reliable source, and it is directly accessible.

How RLB Can Help

RegLeg's published Hallucination Research is available as a free pre-flight reference for Professional Engineers before relying on AI output for regulatory questions. The research catalogues specific failure modes — misquoted thresholds, outdated standards references, fabricated cross-citations — observed across the regulatory instruments that govern engineering practice in multiple jurisdictions. Reviewing findings relevant to a regulation before using AI tools on a matter saves time and reduces the risk of acting on output that sounds authoritative but does not reflect current requirements.

For firms with multiple Professional Engineers working across the same regulatory portfolio, RegLeg offers bespoke regulation deep-dives. These engagements examine the specific instruments your team relies on, map the failure modes most likely to surface in your practice context, and produce a structured reference your engineers can apply consistently. The output is practical and jurisdiction-specific — designed to sit alongside existing compliance workflows rather than replace professional judgement.

RegLeg also develops training material and CPD-aligned content tailored to the failure modes Professional Engineers should actively watch for when AI tools are part of a regulatory research workflow. Where a firm already has an AI-use policy in place, RegLeg can provide a confidential review of that policy against its failure-mode catalogue — identifying gaps, ambiguous guidance, or areas where the policy does not adequately address the known ways AI tools mishandle technical regulatory content.