The FBI’s IC3 recorded $632 million in 2025 investment fraud losses with a reported AI component — out of $893 million in AI-related fraud losses overall — and independent research from identity-verification firm iProov found that people can reliably distinguish real footage from AI-generated fakes only 0.1% of the time. Synthetic CEO endorsements, cloned voice calls, and AI-generated prospectuses are transforming the investment fraud landscape, and the old tells — bad grammar, obvious fakery — are disappearing.

The AI Fraud Toolkit

In 2026, investment fraudsters have access to a sophisticated AI toolkit that was unimaginable five years ago:

  • AI-generated profile photos that pass reverse image searches
  • Real-time voice cloning for phone calls
  • Deepfake video technology for video calls
  • AI-generated investment prospectuses and whitepapers
  • Automated chatbots that maintain convincing conversations at scale

Synthetic Celebrity Endorsements

InvestorWatchdog has identified over 100 investment scam advertisements in 2026 using AI-generated deepfake videos of celebrities and public figures endorsing fraudulent platforms. These videos are increasingly convincing — consistent with iProov’s finding that even trained observers can rarely distinguish them from genuine footage on sight alone.

Voice Cloning: The Toolkit’s Most Underrated Weapon

Video deepfakes get the headlines, but voice cloning may be the more dangerous tool in an investment fraud context, because it’s cheaper to produce and easier to deploy live. Research from McAfee found that as little as three seconds of audio — a snippet pulled from a podcast interview, an earnings call, or a public speech — is enough to produce a convincing clone of someone’s voice.

In investment fraud specifically, this shows up as: a cloned voice of a “senior advisor” following up a video ad with a live phone call, a fake “investor relations” line that sounds exactly like the real firm’s actual staff, or a cloned voicemail confirming a deposit was received. Unlike a video, a cloned voice can improvise in real time to answer questions and overcome objections — which is exactly what makes it effective at the “building confidence” stage of a scam.

A Real Case: The $1.7 Million Fake Elon Musk Video

In a case investigated by CTV News’ W5, a Markham, Ontario woman saw a Facebook video appearing to show Elon Musk promoting a crypto investment opportunity. She sent an initial $250 and was shown a small profit two days later — a pattern designed to build trust before the real request. Over time she was persuaded to send hundreds of thousands more, was shown a fabricated $3 million account balance, and was then told she needed to pay taxes and fees before she could withdraw anything. She ultimately lost $1.7 million, including money borrowed against her home and from family.

A separate investigation by the CBC, based on a leaked internal dataset, found that operations like this typically run as two separate teams working together: one team produces the fake ads and places them on Facebook, Instagram, and X, while a second, unaffiliated call-centre team actually works the victims once they’ve responded. Some of these same networks run a second wave afterward — fraudsters posing as government fraud-recovery agents who contact the same victims claiming they can help retrieve the stolen funds, for a fee. That second wave is a textbook recovery scam — the same victim, targeted twice, by people who already know exactly how much was lost and how vulnerable they are.

AI-Generated Prospectuses

Fraudulent investment prospectuses generated by AI are now often indistinguishable from legitimate documents in language quality and formatting. The content remains fraudulent — but the tell-tale signs of poor grammar and obvious fabrication that previously helped investors identify scams are disappearing fast. Checking the writing quality is no longer a useful test. What still works:

  • Verify named individuals independently — if the document names a law firm, auditor, or fund administrator, contact that firm directly using contact details you find yourself, not details in the document, to confirm the relationship is real.
  • Check for a real registration or ISIN number, and verify it directly on the relevant regulator’s register rather than trusting the number printed in the document.
  • Check the platform’s domain registration date (via WHOIS) — a prospectus describing years of track record attached to a domain registered weeks ago is a clear mismatch.
  • Be suspicious of unfalsifiable claims — proprietary “AI trading algorithms” or “quantitative strategies” that can’t be independently verified are a common wrapper for both AI-generated and traditionally fabricated prospectuses alike.

How to Protect Yourself in the AI Era

  1. Verify celebrity endorsements directly through the celebrity’s official channels — not the video itself, however convincing.
  2. Use AI detection tools for suspicious videos, while knowing they’re imperfect and shouldn’t be your only check.
  3. Verify all investment opportunities through official regulatory registers, regardless of how professional the materials look.
  4. Be especially sceptical of opportunities that came to you — legitimate investments don’t need to cold-call.
  5. Ask for in-person meetings — AI cannot yet replicate physical presence.
  6. Submit a tip to InvestorWatchdog if you spot an AI-generated scam ad or deepfake endorsement — it helps us track and expose these campaigns.

Regulatory Response

Regulators are responding, though not always with new rules. The FCA published a Technology Horizon Scan in late 2026 flagging “synthetic crime” as a major emerging risk — the FCA itself describes this as a foresight document, not formal guidance. The SEC has issued no AI-specific rule; instead, it’s applying existing antifraud and marketing regulations to bring enforcement action against “AI-washing” — six cases since 2024 totalling more than $44 million in alleged fraud. The EU’s AI Act includes transparency requirements for AI-generated and synthetic media, which bear on how deepfake content can be labelled and distributed. Enforcement across all of this remains difficult given the cross-border nature of these operations.

Frequently Asked Questions

How can I tell if a celebrity investment endorsement is a deepfake? Don’t rely on visual judgment alone — research shows people correctly identify deepfakes only a small fraction of the time. Instead, verify the endorsement through the celebrity’s own official social media or website. Genuine celebrities rarely endorse specific trading platforms, and if they did, it would appear on their own verified channels first.

Can AI detection tools reliably spot deepfake videos? They help, but they’re imperfect and can be fooled by newer generation techniques. Treat a detection tool as one signal among several, not a definitive verdict — verification through official channels and regulatory registers remains more reliable.

Has the SEC or FCA banned AI-generated investment content? No. Neither has issued an AI-specific rule for this. They apply existing antifraud and marketing regulations to AI-generated claims and content, meaning the same rules that make a false human claim illegal also apply when AI generates it.

Why don’t scam prospectuses have the grammar mistakes they used to? Because AI language models can now generate polished, professional-sounding documents instantly. The absence of obvious errors used to be a reliable red flag — it no longer is, which is why regulatory verification matters more than document quality when assessing any investment opportunity.

Related Reading