5 unbelievable AI fraud attempts that actually happened in 2026

AI fraud now poses a direct financial and reputational risk to organisations. Criminals can fabricate candidates, executives, evidence and commercial activity with convincing results. Businesses need stronger identity checks, clearer approval processes and far more scepticism about what appears on screen.

Fraud once required forged documents, a plausible story and enough confidence to survive basic questioning.

AI has made that preparation much easier.

AI has made fraud much easier

A fraudster can now produce a polished CV, clone an executive’s voice, manufacture photographic evidence or create thousands of songs without hiring a musician. The results can look credible enough to pass a quick human review or slip through an automated process.

The financial risk is obvious, though the damage can spread further. A fake employee could access customer data. A deepfake executive could persuade a colleague to install malware. A fabricated endorsement could place a trusted institution beside an investment scam.

Five AI fraud cases that have everyone talking in 2026

1. More than 700 suspected North Korean applicants targeted one UK bank

More than 700 suspected North Korean operatives reportedly applied for remote jobs at a major UK bank during one recruitment campaign.

The candidates arrived with polished CVs, strong technical answers and convincing video interviews. Many appeared credible at first glance.

The applications were reportedly flagged by Lisa, an AI recruitment agent created by Alfa AI. It detected repeated language, coordinated behaviour and location inconsistencies across the candidate pool.

The suspected aim went beyond collecting a salary. Securing a remote role could have given an operative access to internal systems, customer information and corporate data.

There is something fitting about an AI recruiter catching hundreds of AI-assisted applicants. Hiring teams should still avoid assuming that software alone will solve the problem. Candidate identity needs to be established through evidence.

2. One man created AI songs and made bots listen to them billions of times

Michael Smith found a strangely efficient business model: create music without musicians, generate listeners without humans and collect royalties in real money.

Smith pleaded guilty to conspiracy to commit wire fraud in March 2026. US prosecutors said he had created hundreds of thousands of songs using AI, then used automated bots to stream them billions of times.

The fake activity was designed to resemble genuine listening and redirect royalty payments away from legitimate artists and rights holders. Smith agreed to forfeit more than $8 million.

The scheme shows what can happen when platforms rely heavily on volume. Once fake content and fake users interact inside the same system, impressive numbers may have little connection to genuine demand.

The same risk applies to traffic, job applications, reviews, enquiries and transactions.

3. Fraudsters created photographs of car crashes that never happened

Faking a motor insurance claim once required someone to damage a vehicle, stage a collision or find a conveniently dented bumper.

Generative AI offers an easier option.

Aviva reported in June 2026 that it had identified more than 18,400 suspected fraudulent claims worth £233 million across its brands during 2025. The insurer said some claimants had used AI-generated accident scenes, manipulated photographs and falsified documents.

A claimant can now invent an accident, create plausible images of the supposed damage and submit them through a digital claims process. No crash is required.

This exposes a weakness in systems built around uploaded evidence. A photograph may support a claim, but its existence no longer proves that the event happened. Insurers need to check image provenance, claim patterns and supporting records.

4. Scammers turned Italy’s central bank governor into an investment influencer

In early 2026, fraudulent videos and articles appeared to show Fabio Panetta, governor of the Bank of Italy, promoting investment products on recognised television programmes.

He had made no such appearances.

The content used deepfake techniques to imitate Panetta and place him inside fabricated broadcasts. The Bank of Italy warned the public and filed a formal complaint with judicial authorities.

The scam borrowed credibility from both a senior public figure and the institution he represented. Someone who would ignore an unknown investment influencer may pay attention when the apparent recommendation comes from a central bank governor.

This creates a reputational problem even when an organisation has done nothing wrong. A convincing fake can circulate widely before the person being impersonated knows it exists.

5. Hackers cloned executives during video calls to install malware

A live video call once felt like strong proof that you were speaking to a real colleague.

That confidence is now risky.

A Financial Times report published in 2026 described how a North Korean hacking group known as PutridSlug had used AI-generated video and audio to impersonate senior executives during Zoom calls in 2025.

Employees were reportedly persuaded to download files or software containing malicious code. The request may have sounded routine, particularly when the apparent executive’s face and voice were visible on screen.

Once installed, the malware could give attackers access to systems and data long after the call ended.

Organisations need separate verification for unusual requests involving software, payments, credentials or sensitive files. A familiar face is no longer enough.

AI identity fraud is a real recruitment threat

These cases share the same weakness: a plausible digital appearance is being mistaken for proof.

That is especially dangerous in recruitment.

McAfee reported that job-related scams increased by more than 1,000% between May and late July 2025, making them one of the fastest-growing fraud categories it tracked during that period.

For hiring teams, the risk is practical. A fake candidate could gain access to internal systems, payroll, customer information, intellectual property and company tools. In a remote role, they may never need to enter an office.

Here's how to outsmart AI job applications

Strengthening background screening is one of the clearest ways to reduce that risk. A polished CV, a convincing video interview and an established LinkedIn profile can all be fabricated. A structured background screening process checks the claims behind the presentation.

A good interview tests judgement, communication and skill. It does not prove identity, employment history or qualifications.

That is where background checks matter. Depending on the role, employers can use identity verification, right to work checks, employment history checks, education verification, criminal record checks and sanctions screening.

The mix should reflect the access and responsibility attached to the job. Veremark’s analysis of almost half a million checks found that employment and education checks identified discrepancies more often than many standard database checks. Employers should consider whether they are checking the information candidates are most likely to misrepresent and whether their screening programme reflects the risks that matter.

AI has made fraud faster, cheaper and more convincing. Employers now need to verify more of what matters before giving someone access, authority or trust.