AI Hallucination Research › Regulators › Major advanced economies › US › CFTC › DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025 › White paper › Detail › Finding
AI Labs · updated 2026-06-03 · methodology v2.1

Finding on 'Q007 Probe' for Claude Sonnet 4.6 with web search ON

RLB Citation ID: RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46
What the RLB Specialist Panel found

Finding on 'Q007 Probe' for Claude Sonnet 4.6 with web search ON

  • Question (paraphrased to protect IP): A futures commission merchant needs to distinguish two '20 percent' requirements in the CFTC's December 2025 digital asset framework — one applied to customer-posted collateral and one to the firm's proprietary holdings. For the customer margin haircut requirement, what specific rule governs the haircut calculation when multiple registered clearing organizations each accept the same digital asset but at different haircut rates?
  • AI's response: "For digital assets not accepted by any registered DCO as initial margin, the FCM must apply a minimum 20% haircut to the current market value of that customer-deposited collateral"
  • Regulator's text: Where multiple DCOs accept the same asset at different haircuts, the FCM must apply the highest such haircut.
  • Why the AI went wrong: The model answered the single-DCO or no-DCO case — the 20 percent floor — but dropped the multi-DCO tie-breaking rule entirely. The question was specifically about the governing rule when multiple clearing organisations each accept the same asset at different rates; the regulator's answer is unambiguous (use the highest). The model's response leaves an FCM believing it may choose among the available DCO haircut rates, when the regulator requires the worst-case selection.
  • Cited source(s):
  • https://financefeeds.com/cftc-issues-faq-on-crypto-collateral-sets-20-charge-... — Pretextual
  • https://www.theblock.co/post/394573/cftc-staff-details-how-crypto-firms-can-u... — Pretextual
Impact for this audience

The dropped multi-DCO worst-case selection rule implicates retrieval-layer answer construction for questions with a numeric threshold and a tie-breaking rule. The model retrieved and stated the base threshold correctly but did not surface the governing rule for the multi-party case, which is the only rule that matters when the question explicitly concerns multiple DCOs accepting the same asset. This is likely a training-data density issue for the FAQ-level elaboration of this rule, combined with a tendency to answer the simpler version of a numeric-threshold question when the more complex governing rule requires an additional retrieval step.

References — raw findings (per AI model)
This finding also affects
← Previous finding Finding on 'Q006 Probe' for Claude Sonnet 4.6 with web search ON
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-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46
Plain text Download 
RegLeg Specialist Panel (2026). "Finding on 'Q007 Probe' for Claude Sonnet 4.6 with web search ON — AI Labs." Citation ID: RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46. RegLegBrief AI Hallucination Research, published 2026-06-03. https://reglegbrief.com/regulators/j3/us/cftc/digital-asset-collateral-tokenized-assets-staff-guidance-2025/whitepaper/finding/US-CFTC-US-001-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-v1-007--sonnet-46-websearch/
APA 7th edition Download 
RegLeg Specialist Panel. (2026). Finding on 'Q007 Probe' for Claude Sonnet 4.6 with web search ON [Hallucination finding RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46]. RegLegBrief AI Hallucination Research. https://reglegbrief.com/regulators/j3/us/cftc/digital-asset-collateral-tokenized-assets-staff-guidance-2025/whitepaper/finding/US-CFTC-US-001-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-v1-007--sonnet-46-websearch/
Bluebook / OSCOLA (US + UK legal) Download 
RegLeg Specialist Panel, Finding on 'Q007 Probe' for Claude Sonnet 4.6 with web search ON [RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46], RegLegBrief AI Hallucination Research (June 03, 2026), https://reglegbrief.com/regulators/j3/us/cftc/digital-asset-collateral-tokenized-assets-staff-guidance-2025/whitepaper/finding/US-CFTC-US-001-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-v1-007--sonnet-46-websearch/.
BibTeX Download 
@misc{reglegbrief_RLB_H_US_CFTC_DIGITAL_ASSET_COLLATERAL_TOKENIZED_ASSETS_STAFF_GUIDANCE_2025_Q007_Sonnet46,
  author    = {RegLeg Specialist Panel},
  title     = {Finding on 'Q007 Probe' for Claude Sonnet 4.6 with web search ON},
  year      = {2026},
  publisher = {RegLegBrief AI Hallucination Research},
  note      = {Hallucination finding Citation ID: RLB-H-US-CFTC-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-Q007-Sonnet46},
  url       = {https://reglegbrief.com/regulators/j3/us/cftc/digital-asset-collateral-tokenized-assets-staff-guidance-2025/whitepaper/finding/US-CFTC-US-001-DIGITAL-ASSET-COLLATERAL-TOKENIZED-ASSETS-STAFF-GUIDANCE-2025-v1-007--sonnet-46-websearch/}
}
← Back to case study summary Case study detail →
About Citation IDs →