FCA Warns AI 'Agentic' Systems and Synthetic Data Could Make Fraud Invisible

The Financial Conduct Authority (FCA) has issued one of its starkest warnings yet about artificial intelligence: autonomous "agentic" AI systems and AI-generated synthetic data could soon make sophisticated financial crime almost impossible for firms to detect using today's controls.

What the FCA is warning about

According to the regulator's latest Emerging Technology Horizon Scan, AI tools can now generate convincing fake "evidence trails" — fabricated audit logs, negotiation emails, shipping manifests and transaction histories — that could conceal Ponzi schemes, bribery or terrorist financing. The FCA has also flagged that "agentic" AI systems, which act semi-independently to execute trades or manage accounts, might commit insider trading, collusion or spoofing without any human directly instructing them to.

The concern is speed. Traditional financial crime controls assume a human is somewhere in the loop, giving compliance teams time to spot and stop suspicious activity. Autonomous, multi-agent AI systems can act far faster, meaning "the velocity of harm ... outpaces traditional firm oversight," according to the regulator's assessment.

The FCA's own response

Rather than simply issuing warnings, the FCA is trying to get ahead of the problem. It has built a fully synthetic dataset containing embedded money-laundering typologies, which firms can use to stress-test their own detection systems without touching real customer data. A second cohort of its Supercharged Sandbox — a controlled environment for testing advanced AI — launched on 13 July 2026. The regulator has cautioned, however, that synthetic datasets risk going stale if they aren't continually updated, meaning firms could end up "optimising their systems for yesterday's risks rather than tomorrow's."

Why this matters if you're not a bank

Most of this sits behind the scenes at regulated firms, but it has knock-on effects for everyday consumers and small businesses:

• Faster-evolving scams. If criminals can use AI to generate more convincing fake documents and identities, expect phishing, romance and investment scams to become harder to spot by eye. The FCA's Warning List and Financial Services Register remain the fastest way to check whether a firm or individual is genuinely authorised before you hand over money.
• More friction, not less. Banks tightening AI-related fraud controls may mean extra verification steps on transfers or account changes — inconvenient, but a sign your bank is taking the threat seriously.
• Business exposure. Small firms using AI tools for bookkeeping, invoicing or customer verification should ask suppliers what safeguards exist against AI-generated fake documents, particularly for anything involving payments.

International comparison

The UK isn't alone in grappling with this. The EU's AI Act, which entered into force in phases from 2024, classifies many financial-crime-relevant AI uses as "high-risk," triggering mandatory risk assessments. In the US, the Treasury's Financial Crimes Enforcement Network (FinCEN) has separately flagged generative AI's role in deepfake-enabled fraud. Australia's financial regulator ASIC has similarly begun consulting on AI governance expectations for licensed firms — all pointing to the same global race between AI-enabled crime and AI-enabled defence.

Checklist: protecting yourself

• Verify any firm asking for money via the FCA Register before transferring funds.
• Be extra sceptical of "too good to be true" investment pitches with polished, professional-looking documentation — AI makes fakes easier to produce.
• Enable extra verification (biometrics, one-time passcodes) wherever your bank offers it.
• If something feels off, report it via Action Fraud or your bank's fraud line immediately.

Key Numbers

• 13 July 2026 — launch date of the FCA's second Supercharged Sandbox cohort
• 2024 — year the EU AI Act began phased entry into force

Sources

• FCA Emerging Technology Horizon Scan 2026 — Herbert Smith Freehills Kramer
• FCA Synthetic Data and Anti-Money Laundering project report — TLT LLP
• FCA sounds alarm as AI reshapes financial crime — Retail Banker International
• FCA Supercharged Sandbox

Educational content only — not financial advice.