The FBI's catalogue of how generative AI gets used in fraud
In December 2024 the FBI published an itemised account of how criminals use generative AI across text, images, audio and video. The most consequential entries are the mundane ones: AI used to correct grammar and spelling for foreign actors targeting US victims, to produce fictitious social media profiles at volume, and to generate the content of fraudulent investment websites.
- Year
- 2024
- Where
- United States
- Outcome
- Ongoing
- Victims
- Not stated in the sources
- Schemes
- Deepfake voice and video scams, Romance scams, Fake investment platforms, Phishing and smishing
- Last reviewed
- 2026-09-06
The facts, as recorded
- IC3 advisory published 3 December 2024.
- The FBI states that generative AI 'reduces the time and effort criminals must expend to deceive their targets' and can correct human errors that would otherwise serve as warning signs.
- AI-generated text is used for social engineering, spear phishing, and romance, investment and other confidence schemes.
- Specifically named: creating voluminous fictitious social media profiles; reaching a wider audience faster; language translation to limit grammatical or spelling errors for foreign actors targeting US victims.
- Also named: generating content for fraudulent cryptocurrency investment websites, and embedding AI-powered chatbots in fraudulent sites to prompt victims to click malicious links.
- AI-generated images are used for believable social media profile photographs and identification documents.
- The FBI notes that creating or distributing synthetic content is not inherently illegal — the concern is its use to facilitate fraud and extortion.
Why this case matters
Public discussion of AI fraud gravitates to the dramatic cases: a cloned voice, a faked executive on a video call. The FBI’s own list is dominated by something much more ordinary, and much more widely consequential.
Grammar. The advisory names, explicitly, the use of AI “to assist with language translations to limit grammatical or spelling errors for foreign criminal actors targeting US victims.”
For thirty years the most widely given piece of anti-fraud advice has been to look for bad English. That was a real signal, and it was a signal about the operator’s location and resources rather than about the scam itself. It has now been removed for the price of a subscription.
The other quiet entries
Fictitious profiles at volume. Romance and investment fraud both need plausible personas. Producing them used to be the constraint on scale.
Fraudulent website content, including for cryptocurrency investment platforms. The polished appearance of a fake trading venue is no longer evidence of a well-resourced operation.
AI-powered chatbots embedded in fraudulent sites, prompting visitors toward malicious links — which means the “support agent” on a fake platform may not be a person at all.
Identification documents. Relevant to SIM swap, where a prosecuted US case turned on a physical fake ID, and to account verification generally.
The FBI’s framing is worth keeping
The advisory says plainly that “the creation or distribution of synthetic content is not inherently illegal”. That matters, because a lot of coverage treats the technology itself as the crime. It is a tool, and the FBI is describing its use in fraud rather than arguing about the tool.
What it tells you about the scheme
The honest summary is that AI has not created a new scam. It has removed several of the cues people were taught to rely on — bad English, an implausible profile photograph, a crudely built website, an unconvincing voice.
What it has not touched is the structure underneath. The money still has to move on a rail that cannot be reversed, the request still has to be verified through a channel you chose, and the shape of the scheme — the stages on every page of this site — is unchanged.
Sources
- Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud. FBI Internet Crime Complaint Center. Accessed 2026-09-06. Supports: Every quoted phrase and itemised use above.