Frontier AI models converted the OECD failing-firm-defence evidentiary list into a closed cumulative test.

Two frontier AI subjects tested by the RLB Specialist Panel converted Section III.11.b's 'inter alia' evidentiary criteria into a closed three-condition cumulative test with all-or-nothing pass/fail framing and substituted an inevitability-of-exit counterfactual for the Recommendation's comparative-harm third limb.

The pattern in one line

Frontier AI models tested on the 2025 OECD Merger Review Recommendation converted the failing firm defence's 'inter alia' evidentiary criteria into a closed three-condition cumulative test and mischaracterised the operative third limb, producing legal answers that would under-prepare corporate banking firms on additional evidentiary lines an authority may require.

How the RLB Specialist Panel tested this

The questions in this cell were prepared by the RLB Specialist Panel based on real, practical AI usage in the workflows that legal teams at corporate banking firms actually use AI for under the 2025 OECD Merger Review Recommendation. Each question targets a specific deliverable type where an AI assistant is plausibly the first draft: an operative-section summary in a client memo, a remedies-hierarchy paragraph in a regulatory-strategy paper, a failing-firm-defence formulation in a transaction-committee briefing, a Council-reporting cadence line in an inter-agency engagement note. The Panel issued each question to two frontier AI subjects with web search active.

The Panel then bound every AI response to verbatim regulator-issued source text held as primary substrate, comparing the model output against the operative OECD Recommendation text and the supporting OECD guideline used to verify cross-references. Only responses where the AI subject was demonstrably wrong against the verbatim regulator-issued source text are published as findings; responses that were substantively correct, or that refused on calibration grounds, are retained internally and not surfaced.

What the models got wrong

Finding: 'Inter alia' evidentiary criteria converted into a closed cumulative test; third condition mischaracterised. The Specialist Panel asked, in application form, what conditions a merging party must satisfy to invoke the failing firm defence under the 2025 OECD Merger Review Recommendation, and how demanding the overall standard is. Claude Opus 4.7 with web search active answered that the defence requires three cumulative conditions, the third being an absent-the-merger counterfactual under which the target's productive assets would inevitably exit the market (RLB-H-INT-OECD-OECD-MERGER-REVIEW-RECOMMENDATION-2025-Q005-Opus47).

Claude Sonnet 4.6 with web search active produced a parallel closed-test framing: three conditions, all of which must be satisfied simultaneously, failure on any one being fatal to the defence (RLB-H-INT-OECD-OECD-MERGER-REVIEW-RECOMMENDATION-2025-Q005-Sonnet46).

The substrate held by the Panel records Section III.11.b verbatim: authorities should 'require, inter alia, evidence that the business was likely to have exited and the existence of less anti-competitive alternative buyers or other options for reorganisation are not viable, and that the exit of the firm's assets would cause more harm to competition than the merger.' 'Inter alia' signals a non-exhaustive list; the Recommendation does not impose a closed three-condition cumulative test, and the operative third limb is a comparative-harm test (asset exit harm versus merger harm), not an inevitability-of-exit counterfactual.

Why this matters for Legal teams at Corporate Banking firms

For legal teams at corporate banking firms advising on cross-border merger transactions touching the 2025 OECD Merger Review Recommendation, citation accuracy on the operative architecture, on Section IV.3 remedies hierarchy, and on Section III.11.b failing firm defence is load-bearing in every authority-facing submission, every board memo, and every transactional document. A counterparty or competition authority who identifies a structural inflation, a misattributed sub-hierarchy, or a closed-cumulative-test framing on first reading calls the entire piece of advice into question.

The structural-architecture failure is the most directly visible: a board memo or regulator-facing submission that lists 'international co-operation' or 'monitoring' as operative RECOMMENDS sections is wrong on first reading. The Section IV.3 EU sub-hierarchy import is the most insidious failure, reading as authoritative because the EU framework is real, but presenting EU practice as OECD content imports the wrong normative baseline into the firm's remedy strategy.

The regulator's actual position

'Inter alia' criteria with a comparative-harm third limb; no closed cumulative test. Section III.11.b of the 2025 Recommendation directs authorities to 'require, inter alia, evidence that the business was likely to have exited and the existence of less anti-competitive alternative buyers or other options for reorganisation are not viable, and that the exit of the firm's assets would cause more harm to competition than the merger.' Two textual signals matter. First, 'inter alia' marks the criteria as non-exhaustive: authorities may require additional evidentiary lines on the facts.

Second, the third limb is a comparative-harm test, the loss of competition from asset exit must be more harmful than the merger, not an inevitability-of-exit counterfactual that asks whether the assets would leave the market. The Recommendation neither numbers the conditions nor frames them as a closed all-or-nothing cumulative test.

What this tells us about AI for Legal teams at Corporate Banking firms

For legal teams at corporate banking firms working with AI on the 2025 OECD Merger Review Recommendation, the recurring pattern is an Inter-Alia-to-Closed-Test Conversion: the AI subjects collapsed Section III.11.b's non-exhaustive evidentiary criteria into a closed three-condition cumulative test and substituted an inevitability-of-exit counterfactual for the operative comparative-harm third limb. The downstream consequence is a failing firm defence submission, or a memo on the standard, that under-prepares on additional evidentiary lines the authority may require under 'inter alia' and that mischaracterises the third limb.

The defensive workflow is a substantive read of Section III.11.b before any failing firm defence paragraph enters a client deliverable.

What the RLB Specialist Panel is doing about it

The RLB Specialist Panel is engaging with the AI subjects' developers and with practitioner audiences working under the 2025 OECD Merger Review Recommendation. The Panel maintains an audit register of confirmed hallucinations bound to verbatim regulator-issued source text, surfaces them on the live regulation page and on each audience-specific briefing, and accepts right-of-reply submissions from the AI subjects' developers and from regulator-side reviewers.

For legal teams at corporate banking firms this means the same questions can be re-issued against successor model releases; the bound substrate makes it straightforward to verify whether a specific failure mode has been corrected upstream, or whether the same hallucination is still being produced. Partnership briefings with AI labs are offered against the audit register, not against synthesised demonstrations, so the corrections that matter are evidenced against the operative OECD text rather than against a paraphrase chain.

What Legal teams at Corporate Banking firms teams should do

For legal teams at corporate banking firms drawing on AI in workflows that touch the 2025 OECD Merger Review Recommendation, the practical action items are direct:

• Anchor every operative-section reference, every remedies-hierarchy paraphrase, and every failing-firm-defence formulation produced by an AI assistant against the verbatim OECD Recommendation text and the supporting OECD guideline cross-references, before the citation enters a memo, opinion, or regulatory submission.
• Treat any AI-generated structural summary of the 2025 Recommendation as a high-risk paragraph: read the operative RECOMMENDS block directly and confirm that Section V (ex-post assessment) is reflected, and that international co-operation and monitoring are not presented as standalone operative sections.
• When advising on Section IV.3 remedies hierarchy, validate that any structural-remedy sub-ordering presented as OECD content is in fact OECD text and not an EU Commission practice convention imported by paraphrase.
• When advising on the failing firm defence under Section III.11.b, preserve the 'inter alia' signal and the comparative-harm third limb in the deliverable; do not let an AI-drafted closed cumulative test enter the client work product unmodified.
• Maintain a firm-level AI-assist log that records every AI-drafted OECD Recommendation citation entering a client or regulator-facing deliverable, with the verification step explicitly recorded for post-hoc audit.
• Subscribe to the RLB Specialist Panel audit register on the 2025 OECD Merger Review Recommendation to receive notifications when a confirmed AI hallucination is added, withdrawn, or corrected against successor model releases.

Right of Reply

These findings and associated work have been put up in public with a view of the greater good for the development of a safer AI ecosystem. Any party reading this or any finding on reglegbrief.com may contact us and have an unconditional right of reply; the Specialist Panel will publish any factual correction or contextual response alongside the original finding, with no editorial gatekeeping. Researchers, regulators, and compliance teams with questions on methodology or specific findings can reach the Specialist Panel via the same channel.

Source & Methodology Standards

RegLeg Brief is operated by Verdus Technologies Pte. Ltd. (UEN 201616982R), incorporated in Singapore. The RLB Specialist Panel, with an aggregate of over 60 years of public-policy and industry experience, documents only confirmed hallucination findings, under a methodology that requires a verbatim regulator excerpt for every documented claim. All findings, citation IDs, model outputs, regulator excerpts, and methodology notes are open-access.

Primary source verified: OECD/LEGAL/0333, Recommendation of the Council on Merger Review (2025 Revision) · Substrate documents: R1-REGULATION-00001 · OECD portal: oecd.org/legal

Citation IDs referenced:

• RLB-H-INT-OECD-OECD-MERGER-REVIEW-RECOMMENDATION-2025-Q005-Opus47