AI-Generated Fraud Will Break Standard Due Diligence (and What to Do About It)

When everything on paper checks out but it’s all synthetic.

The vendor file was flawless:

Clean company registry. Signed vendor contracts. Escrow confirmations.

Two quiet calls later, the deal collapsed.

What made this one different?

AI didn’t create the fraud but it made the fake story faster, cleaner, and harder to detect. And this is only the beginning.

The new face of fraud

Large language models and generative AI can now:

• Draft “clean” contracts, term sheets, and letters that pass legal review

• Build synthetic identities and fake work histories that survive KYC

• Spin up vendor websites, filings, and references in hours

On paper, everything lines up. In reality, nothing does.

Why standard due diligence will fail

Most due diligence still assumes documents are ground truth. That’s broken in three places:

1. Document-centric checks – If all sources lead back to the same fabricated paperwork, you’re validating the illusion.

2. Single-source verification – Calling the number on the letterhead only confirms the letterhead.

3. Static screening – Lists and databases miss fast-moving synthetic networks.

A modern playbook for synthetic fraud

Here’s what works in the field:

• Provenance over presentation – Verify where a document came from, who touched it, and when not just how it looks.

• Cross-source contradictions – Correlate people, entities, and timelines across independent sources (regulators, trade data, court records, industry backchannels).

• Identity beyond KYC Behavioural and reputational footprints (past deals, counterparties, litigation patterns) are harder to fake than IDs.

• Network analysis – Map relationships; fraud clusters reuse the same nominees, addresses, and advisors.

• Human backchannel – Quiet reference calls with people not listed in the file almost always surface what the paperwork hides.

An anonymised field example

A cross-border deal looked flawless: clean company registry, signed vendor contracts, escrow confirmations. Two things didn’t smell right: (1) directors with impressive CVs but thin industry footprints, and (2) an “independent” introducer who kept reappearing in unrelated transactions.

Cross-source checks showed recycled addresses, a dissolved auditor resurfacing under a new brand, and a reference letter hosted on a domain registered last month. On paper: perfect. Off paper: walk away.

Questions every board should ask this quarter

• Provenance – How do we verify origin and handling of documents, not just their content?

• Independence – Which checks rely on data controlled by the counterparty?

• Red-team – Who is incentivised to break our story before we sign?

Why this matters now

AI won’t replace due diligence. It will supercharge both good actors and bad ones. The gap will be defined by who upgrades their process first.

If you want an extra set of eyes on a deal, vendor, or prospective hire, I run Second Opinion Check. We specialise in human-verified, cross-border vetting that looks beyond paperwork to people, patterns, and provenance.

By Titus Van Heur