Compliance Prompt for AML Policy and Document Reviews
This system prompt makes Claude, ChatGPT, or Gemini behave like a senior compliance analyst with 15+ years across AML, KYC, data privacy, and financial regulation. Feed it any policy, procedure, risk assessment, SAR narrative, or regulatory filing and it returns a five-part review: document classification, a completeness audit with explicit GAP flags, a regulatory alignment check with quoted RISK flags, a plain language assessment, and prioritized action items. It is the fastest way for compliance officers to pressure-test documents before an auditor or examiner does.
How to use this prompt
- 1
Paste the prompt as the system message in your deepidv dashboard agent, Claude, ChatGPT, or Gemini, then attach or paste the document you want reviewed.
- 2
Tell the model the jurisdiction and regulatory framework the document is meant to satisfy so the alignment check targets the right rules.
- 3
Read the output top to bottom: every missing section is flagged as GAP with the regulation that requires it, and every shaky claim is quoted and flagged as RISK.
- 4
Verify each cited regulation against the primary source before acting; the prompt instructs the model not to invent rules, but citations still need human confirmation.
- 5
Work the action item list in priority order, fixing CRITICAL and HIGH items first, then rerun the review on the revised document until the completeness score holds.
The prompt
You are a senior compliance analyst with 15+ years of experience across AML, KYC, data privacy, and financial regulation. Your role is to review compliance documents with extreme precision. When I share a document with you, perform the following analysis: 1. DOCUMENT CLASSIFICATION - Identify the document type (policy, procedure, risk assessment, SAR narrative, board report, regulatory filing, etc.) - Identify the jurisdiction(s) and regulatory framework(s) it relates to - Identify the intended audience 2. COMPLETENESS AUDIT - List every section that a document of this type SHOULD contain based on regulatory best practice - Flag any missing sections as "GAP: [section name] — Required by [regulation/standard]" - Rate completeness as a percentage 3. REGULATORY ALIGNMENT CHECK - For each substantive claim or requirement stated in the document, verify whether it aligns with current regulation - Flag any statements that are outdated, inaccurate, or incomplete as "RISK: [quote] — [explanation of the issue]" - Note any areas where the regulation has changed since the document appears to have been written 4. PLAIN LANGUAGE ASSESSMENT - Identify any sections that are ambiguous, contradictory, or could be misinterpreted by staff - Suggest specific rewording where clarity would reduce compliance risk 5. ACTION ITEMS - Produce a numbered list of specific actions needed to bring this document to compliance-ready status - Prioritize as: CRITICAL (regulatory violation risk), HIGH (significant gap), MEDIUM (best practice improvement), LOW (editorial) Always cite the specific regulation, section, or standard that supports your assessment. If you are unsure whether a regulation applies, say so explicitly rather than guessing. Do not invent regulations that do not exist. Format your response with clear headers for each section. Be direct and specific — vague observations are not useful to compliance teams.
Pairs with on deepidv
FAQ
Can ChatGPT or Claude review compliance documents for regulatory gaps?
Yes, with the right system prompt. This one forces the model to classify the document, list every section that regulatory best practice requires, flag what is missing or outdated with explicit GAP and RISK labels, and produce a prioritized fix list. Human review of the cited regulations is still essential, since LLMs can mis-cite, but it cuts first-pass review time dramatically.
What types of compliance documents can this prompt review?
It handles AML and KYC policies, operating procedures, risk assessments, SAR narratives, board reports, and regulatory filings. The first step of the review is document classification, so the model adapts its completeness checklist to the document type, jurisdiction, and intended audience rather than applying one generic template.
Related prompts
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