FinTechReview Prompt

Review Prompt to Map Executive Deepfake Impersonation Exposure and Ceremony Gaps

This Luna review prompt treats the Fideuram chain, a spoofed message plus one cloned voice moving nine figures, as the template attack and audits your leadership against it. Luna, the deepidv compliance agent, builds a clone-surface inventory grading each senior executive's and key counterparty's publicly scrape-able voice and video (earnings calls, panels, podcasts, platform video) as impersonation training data, then draws an authority intersection map crossing who can instruct or approve payments and data releases above each threshold with the channels they habitually use, flagging every authority-holder reachable by consumer messaging as a priority target. A ceremony gap audit traces what stands between a convincing instruction and execution (enrolled-channel rules, directory callbacks, identity-anchored approval, dual authorization, action binding) and names each missing stage as a finding, a pre-commitment check confirms executives have signed the no-exemption statement and enrolled in verification, and a drill and insurance readiness review tests whether your crime and cyber policies' social-engineering conditions match the ceremony you actually record. Everything lands in a ranked exposure register by executive, channel, authority, missing control, owner, and date. Built for CFOs, treasury, and security leads who can no longer treat a recognizable voice as a control.

How to use this prompt

1
Open Luna in the deepidv dashboard and paste the full prompt, or run it in Claude, ChatGPT, or Gemini if you are drafting the exposure register outside the platform.
2
Add an INPUT block below the prompt listing your senior executives and key counterparties (counsel, auditors, bankers), their payment and data-release authority thresholds, the channels they use for business requests, and the approval ceremony that runs today.
3
Run the prompt and read the authority intersection map first: every authority-holder reachable by consumer messaging is a priority target, because that is the channel the Fideuram chain opened with.
4
Work the ranked exposure register top down, assigning an owner and date to each missing control, and route the pre-commitment gaps to the executives themselves so the no-exemption statement gets signed and verification enrollment gets done.
5
Re-run after leadership or authority changes and at each crime and cyber policy renewal, and pair it with a simulated-impersonation drill so the register reflects tested ceremony rather than assumed ceremony.

The prompt

Luna, run an executive impersonation exposure review, treating the Fideuram chain (spoofed messaging plus one cloned voice moving nine figures) as the template attack.

Produce:
1. Clone-surface inventory: per senior executive and key counterparty (counsel, auditors, bankers), the publicly scrape-able voice and video corpus, earnings calls, panels, podcasts, platform video, graded by volume and quality as impersonation training data.
2. Authority intersection map: who can instruct or approve payments and data releases above each threshold, crossed with the channels those people habitually use for business requests, flagging every authority-holder reachable by consumer messaging as a priority target.
3. Ceremony gap audit: for the top authority intersections, trace what currently stands between a convincing instruction and execution: enrolled-channel rules, directory callbacks, identity-anchored approval, dual authorization, action binding, and name each missing stage as a finding.
4. Pre-commitment status: whether executives have signed the no-exemption statement (no genuine instruction will ever require bypassing verification) and enrolled in verification so the secure path is also the fast path.
5. Drill and insurance readiness: last simulated-impersonation drill date and findings, plus whether our crime and cyber policies' social-engineering coverage conditions match our actual recorded ceremony.

Output a ranked exposure register: executive, channel, authority, missing control, owner, date. Treat "everyone would recognize his voice" as an exposure, not a control.

Test it in Claude or another LLM

This prompt is built for Luna inside deepidv, where Luna works from your live verification records and per-decision evidence. You can dry-run the exposure review in any general LLM first with a synthetic leadership roster and approval ceremony before pointing it at real executives.

1
Paste the full prompt into Claude, ChatGPT, or Gemini, replacing the opening 'Luna,' with a role instruction such as 'Act as an executive impersonation risk auditor mapping a leadership team against a cloned-voice payment fraud.' Keep the five numbered deliverables as written.
2
Below the prompt, paste the synthetic sample block from this page so the model has executives, counterparties, authorities, channels, and a ceremony to audit.
3
Add one framing line: 'This is synthetic test data. Where a control is not described in the input, treat it as missing rather than assuming it exists.'
4
Check the output shape: a clone-surface inventory, an authority intersection map, a ceremony gap audit, a pre-commitment status, and a drill and insurance readiness review, rolled into a ranked exposure register with an owner and date per row. If any executive is credited with a control the input did not name, tighten the framing and re-run.
5
Once the shape is right, run it live in the deepidv dashboard where Luna works from your real verification records.

Synthetic sample data to paste alongside the prompt

Fake test data, safe to share with any LLM. Swap in your own once the output looks right.

EXECUTIVES (synthetic, fake):
- CEO A. Example: 40+ hours of public earnings calls and podcasts, approves payments above $5M (fake threshold), takes business requests on WhatsApp
- CFO B. Sample: two conference panels online, releases wires above $1M, email and desk phone only
COUNTERPARTIES (fake): outside counsel C. Test, frequent podcast guest, can instruct on deal payments
CEREMONY TODAY (fake): callback to the number in the request, single approver below $10M, no action binding
PRE-COMMITMENT (fake): no executive has signed a no-exemption statement
INSURANCE (fake): crime policy social-engineering cover requires documented callback verification
LAST DRILL (fake): none on record

Pairs with on deepidv

Sources & further reading

FAQ

What is the Fideuram impersonation chain?

It is the attack pattern behind the case in which Fideuram, the private banking arm of Intesa Sanpaolo, lost about 95 million euros: a WhatsApp message to the then-chair appearing to come from the group chief executive, followed by a call in an AI-cloned voice imitating a senior lawyer, was enough to get transfers to foreign accounts arranged. No systems were compromised and no documents were forged, which is why the prompt treats it as a ceremony failure rather than a technology failure.

What does the Executive Clone Exposure Review produce?

A clone-surface inventory, an authority intersection map, a ceremony gap audit, a pre-commitment status check, and a drill and insurance readiness review, consolidated into one ranked exposure register by executive, channel, authority, missing control, owner, and date. Every missing verification stage between a convincing instruction and execution comes back as a named finding.

Why does a recognizable voice not count as a control?

Because an executive's public voice and video corpus (earnings calls, panels, podcasts) is training data for a clone, so recognizing the voice only proves the attacker had enough samples. The prompt treats "everyone would recognize his voice" as an exposure and looks instead for controls a clone cannot pass: directory callbacks, identity-anchored approval, dual authorization, and action binding.

Run it with live verification data

These prompts work in any LLM. Inside the deepidv dashboard, Luna, Arbiter, and Arc run them against your real sessions, screening lists, and audit trails.

Book a Demo