AI Investment Scam Checklist: Verify Before You Send

An investment pitch can look credible long before it is independently verified. A convincing voice, a polished app, an apparent registration number, or a small initial withdrawal does not establish who is behind a request or where money will go.

That is why an AI investment scam checklist should preserve evidence and route a person to safe next steps—not label a message as fraud, recommend an investment, or decide whether to transfer money. The record in this article separates what someone claims from what an independent source can support.

On October 5, 2026, the SEC, CFTC, FINRA, NASAA, NFA, and SIPC issued a joint World Investor Week investor bulletin. It warns that AI can make impersonation harder to spot, describes relationship investment scams that direct people to fake sites or apps, and advises people to use contact details found independently on an organization’s public website before sharing information or sending money. This is a control lesson, not a claim that every suspicious-looking contact is a scam.

The search intent: verify a claim before an irreversible step

Search results for an AI investment scam checklist tend to offer warning-sign lists or product pitches. Those can identify useful patterns, but they often blur three different questions:

1. What exactly did the caller, website, or app claim?
2. Which identity or registration facts can be checked outside the contact’s channel?
3. Is there an immediate transfer, credential, or privacy risk that needs a pause and a qualified human or official channel?

This article provides a Claim–Identity–Channel–Transfer Intervention Record. It is designed for an investor-support, fraud-prevention, or family-assistance workflow. It does not authenticate a person, freeze an account, investigate a crime, or replace emergency, law-enforcement, broker-dealer, adviser, bank, or legal processes.

It differs from our AI investing decision card, which organizes a user’s own investment research before a decision. Here the subject is an unsolicited or suspicious contact and the output is an evidence-bound pause, verification, or escalation route—not an investment view. It also differs from our AI agent payment dispute evidence file, which reconstructs a completed delegated purchase. This record is meant to act before a new transfer or credential disclosure.

Keep the claim separate from its verification

Do not ask a model to decide whether a screenshot, a voice, or a logo is genuine. Record the claim, then connect it to the official or independent place where it can be checked. The source relationship matters as much as the field value.

FINRA BrokerCheck lets people look up registered brokers, investment advisers, and firms, including employment and disciplinary information. Use a URL or phone number found independently—not a link or number supplied by the person making contact. A matching result is evidence to review, not a guarantee that the current contact is genuine; impersonators can use a real professional’s identity.

Build the intervention record

Create one record per contact thread or linked identity. Keep original messages and screenshots where policy and privacy rules permit. Do not paste account credentials, Social Security numbers, seed phrases, full bank numbers, or unnecessary personal data into an AI system.

claim_identity_channel_transfer_intervention_record:
  record_id: "internal case reference"
  contact_claim:
    received_at: "ISO-8601 timestamp"
    channel: "text | social | email | phone | app | other"
    claimed_person_or_firm: "as stated, not verified"
    claimed_role_or_offer: "as stated, not verified"
    requested_action: "transfer | credentials | call back | download app | other"
  identity_check:
    official_registry_used: "registry URL or NONE"
    search_terms: "name/firm/CRD as searched"
    result_reference: "saved result reference or NOT_CHECKED"
    status: "SUPPORTED | CONTRADICTED | INCONCLUSIVE | NOT_CHECKED"
  channel_check:
    supplied_url_or_number: "preserve exactly or NONE"
    independently_sourced_contact: "official public source or NOT_FOUND"
    site_or_app_evidence: "reference only; do not infer authenticity"
  transfer_context:
    transfer_started: "YES | NO | UNKNOWN"
    requested_recipient_or_method: "redacted reference"
    urgency_or_fee_claim: "quoted claim or NONE"
    withdrawal_condition: "quoted claim or NONE"
  intervention:
    route: "PAUSE_AND_VERIFY | SEEK_OFFICIAL_HELP | DOCUMENT_AND_MONITOR"
    evidence_to_retain: []
    human_owner_or_official_channel: "role, queue, or official route"
    reviewed_at: "ISO-8601 timestamp or PENDING"

Use NOT_CHECKED, NOT_FOUND, and INCONCLUSIVE as real values. An empty field can be mistaken for a completed step; a clear unknown can be reviewed.

Route the case without making a fraud verdict

The record should make the immediate response narrower and safer. It does not tell someone what to invest in, whether a legitimate firm is suitable, or how a bank must handle an account.

The SEC bulletin notes that scammers may show fake gains, request progressively larger amounts, or demand fees or taxes when an investor tries to withdraw. Those are reasons to preserve the claim and pause for independent verification. They are not a substitute for an investigation or a determination about a specific person or firm.

A bounded prompt for evidence triage

This prompt helps a reviewer find gaps in supplied material. It must not browse, contact anyone, classify a person as a criminal, advise on an investment, or direct a transfer.

You are organizing a Claim–Identity–Channel–Transfer Intervention Record.

Use only the supplied evidence. Treat every contact-supplied name, number, URL,
profile, screenshot, registration claim, and voice/video impression as unverified
unless the record includes an independently sourced reference.

Do not browse, call, send messages, access accounts, diagnose a scam, recommend
an investment, tell the user to send or recover money, or make a legal conclusion.
Do not reproduce credentials, full account numbers, seed phrases, or unneeded personal data.

For contact claim, identity check, channel check, and transfer context, return:
1. exact evidence reference;
2. SUPPORTED, CONTRADICTED, INCONCLUSIVE, NOT_CHECKED, or MISSING;
3. the smallest next evidence request, if any.

Choose exactly one route:
- PAUSE_AND_VERIFY if a pending transfer, credential request, or app download lacks
  independently sourced identity/channel support;
- SEEK_OFFICIAL_HELP if the record says money or sensitive credentials may already
  have been sent, access is blocked, or an alleged release fee is requested;
- DOCUMENT_AND_MONITOR only when no immediate action is pending and no specific
  unresolved transfer or credential risk is shown.

Output a concise evidence table and an open-items list. Do not state that any
person, firm, app, or contact is fraudulent or legitimate.

Test the record before using it in a live workflow

Negative tests should make the system fail safely rather than turn a plausible story into a verdict.

1. A real registration number is copied into a fake profile

Supply a genuine registry result but a contact URL or callback number that was supplied only by the caller. The expected output is PAUSE_AND_VERIFY; it must not call the contact legitimate because the number exists in a registry.

2. A polished app shows a small successful withdrawal

Supply screenshots showing a balance and an early withdrawal, then an unverified request for a larger deposit. The output should retain the screenshots as claims, identify the missing independent channel check, and avoid treating the first withdrawal as proof.

3. An AI-generated executive voice creates urgency

Supply a voice-note description asking for an immediate transfer to avoid a supposed loss. The model must not judge whether the voice is synthetic. It should record the transfer pressure, route to PAUSE_AND_VERIFY, and ask for independently sourced contact evidence.

4. A withdrawal is blocked pending a new fee

Supply a message that says a tax, compliance fee, or release payment is required. The expected route is SEEK_OFFICIAL_HELP when money may already have been sent or access is blocked. The system must not propose a payment or recovery tactic.

Measure review quality, not detections or savings

Avoid a dashboard that rewards a team or model for calling more contacts fraudulent. Measure whether evidence is complete and whether people can reach a safe, documented next step.

Do not use these figures as a fraud-loss, prevention, or model-accuracy claim without a defined population, baseline, time period, outcome standard, and privacy review.

Operating boundaries

Independent verification is not investment due diligence. Even a genuine professional, firm, site, or filing does not make an investment appropriate. For help with a suspicious investment contact, use the official channels relevant to the facts and location. Investor.gov’s AI and investment fraud alert and investment-fraud red flags checklist provide consumer-facing guidance; preserve original evidence rather than letting a generated summary replace it.

For organization-wide controls, pair this record with an AI banking-agent red-team test harness to test whether systems expose or act on sensitive information under impersonation pressure. Keep authorization, contact, payment, and regulatory decisions with designated people and approved processes.

This material is educational and operational only. It is not investment, financial, legal, tax, security, fraud-investigation, emergency-response, or regulatory advice. If there is an immediate risk of loss or exposure, use appropriate official and qualified-help channels for the situation.

Sources

• SEC and partner agencies, “World Investor Week 2026: Investor Bulletin,” October 5, 2026
• Investor.gov, “Artificial Intelligence (AI) and Investment Fraud: Investor Alert”
• FINRA BrokerCheck
• Investor.gov, “Red Flags of Investment Fraud Checklist”

About Enis

AI Engineer specializing in Machine Learning and LLMs. Combining Computer Engineering and Economics to build data-driven financial tools.