AI-Generated Fraud Could Hit $40 Billion by 2027. Standards Bodies Just Announced New Tools to Fight It

The Problem: Nobody Can Tell What's Real Anymore

You can't trust your eyes anymore, and that's not a metaphor. Photos, videos, and audio generated by AI have gotten good enough that regular people, journalists, and even law enforcement can struggle to separate the real from the fake.

"You now don't know what if something is really fake or not," said Touradj Ebrahimi, a professor at the Swiss Federal Institute of Technology, according to ZDNET. Ebrahimi is leading efforts to fix that problem, working with three of the biggest names in global standards-setting: the International Electrotechnical Commission (IEC), the International Organization for Standardization (ISO), and the International Telecommunication Union (ITU).

The money angle deserves attention. Deloitte's Center for Financial Services estimates generative AI could push US fraud losses to $40 billion by 2027, according to ZDNET's reporting. That's up from $12.3 billion in 2023, a more than 3x jump in four years driven largely by AI's ability to fake images, voices, and documents at scale.

What Just Got Announced

At the AI for Good conference in Geneva, held under the United Nations, the IEC and ISO introduced two new additions to their JPEG Trust standard, according to ZDNET. The original JPEG Trust standard came out last year and was built to embed metadata directly into JPEG files, essentially a digital paper trail showing where an image came from and whether it's been altered.

The new additions are still in progress. The point is simple: give end users and companies a way to check an image's history and authenticity without relying on guesswork.

The Catch Nobody Should Ignore

The JPEG Trust standard doesn't stop anyone from creating fake content. It doesn't label AI-generated images at the moment they're made. It's a verification tool for the back end, not a prevention tool for the front end.

"Fraudsters will not label their content as AI," Ebrahimi said, according to ZDNET. "If somebody wants to break the law, they're not going to break the law and follow the other law that says that content needs to be labeled."

A voluntary labeling standard only works on people who were never trying to deceive anyone in the first place. Criminal fraudsters, state-backed disinformation operations, and scammers running romance or investment schemes aren't going to add a metadata tag to their fake content just because ISO asked nicely.

The actual value lies in giving legitimate platforms, news organizations, and verification services a shared technical language to check content that claims to be authentic. If a photo lacks proper trust metadata, or the metadata has been stripped or tampered with, that's a red flag worth investigating. It's a forensic tool, not a firewall.

Who's Actually Going to Use This

Standards are only as good as adoption. ISO and IEC can publish all the specifications they want, but if camera manufacturers, social media platforms, and news organizations don't build support for JPEG Trust into their products, it's a document sitting on a shelf.

The framework's success will depend on whether major players—think Adobe, Google, Apple, and the big social platforms—actually implement it at scale. None of those companies' specific adoption plans were detailed in the available reporting, which leaves a real open question about timeline and enforcement.

There's also a bigger structural issue: verification tools work best when there's a chain of custody from the moment content is captured. Retrofitting trust onto the billions of images already floating around the internet, most without any embedded metadata at all, is a much harder problem than tagging new content going forward.

What's Next

The two new JPEG Trust additions are still being finalized, and no firm release date was specified in available reporting. The bigger test will come once these standards move from committee rooms in Geneva to actual products people use every day.

The $40 billion fraud estimate from Deloitte stands as a warning, not a settled outcome. Whether that number gets worse or better between now and 2027 will depend less on what standards bodies write down and more on whether tech companies, banks, and law enforcement actually build and enforce the tools to catch fakes before they cause damage.