Sam Altman not too long ago warned that AI-powered fraud is coming “very quickly,” and it’ll break the methods we depend on to confirm id.
It’s already occurring and it’s not simply coming for banks; it’s hitting each a part of our authorities proper now.
Each week, AI-generated fraud is siphoning tens of millions from public profit methods, catastrophe aid funds, and unemployment applications. Felony networks are already utilizing deepfakes, artificial identities, and enormous language fashions to outpace outdated fraud defenses, together with simply spoofed, single-layer instruments like facial recognition, and so they’re profitable.
We noticed a glimpse of this in the course of the pandemic, when fraud rings exploited gaps in state methods to steal a whole lot of billions in unemployment advantages. It wasn’t simply individuals sporting masks to bypass facial recognition. It was AI-generated pretend identities, voice clones, and solid paperwork overwhelming methods that weren’t constructed to detect them. In the present day, these techniques are extra superior, and totally automated.
I work with over 9,000 companies throughout the nation. As I testified earlier than the U.S. Home of Representatives twice this yr, what we’re seeing within the subject is evident. Fraud is quicker, cheaper, and extra scalable than ever earlier than. Organized crime teams, each home and transnational, are utilizing generative AI to imitate identities, generate artificial documentation, and flood our methods with fraudulent claims. They’re not simply stealing from the federal government; they’re stealing from the American individuals.
The Small Enterprise Administration Inspector Common now estimates that almost $200 billion was stolen from pandemic-era unemployment insurance coverage applications, making it one of many largest fraud losses in U.S. historical past. Medicaid, IRS, TANF, CHIP, and catastrophe aid applications face related vulnerabilities. Now we have additionally seen this firsthand in our work alongside the U.S. Secret Service defending the USDA SNAP program, which has turn into a buffet for fraudsters with billions stolen nationwide each month. The truth is, in a single day utilizing AI, one fraud ring can file tens of hundreds of pretend claims throughout a number of states, most of which can be processed mechanically until flagged.
We’ve reached a turning level. As AI continues to evolve, the size and class of those assaults will enhance quickly. Simply as Moore’s Legislation predicted that computing energy would double each two years, we’re now dwelling by means of a new form of exponential progress. Gordon Moore, Intel’s co-founder, initially described the pattern in 1965, and it has guided many years of innovation. I consider we could quickly acknowledge an identical precept for AI that I name “Altman’s Legislation”: each 180 days, AI capabilities double.
If we don’t modernize our defenses with the identical tempo as technological developments, we’ll be completely outmatched.
What we desperately want is smarter instruments and infrastructure, no more forms.
Meaning layering superior id verification, not simply facial scans or passwords. It means utilizing real-time information, behavioral analytics, and cross-jurisdictional instruments that may flag anomalies earlier than cash goes out the door. It additionally means reviving what has already labored: instruments just like the Nationwide Accuracy Clearinghouse, which flagged billions of {dollars} in duplicate profit claims throughout state traces earlier than it was shut down.
AI is a power multiplier, however it may be weaponized extra simply than it may be wielded for defense. Proper now, criminals are utilizing it higher than we’re. Till that adjustments, our most weak methods and the individuals who rely on them will stay uncovered.
The opinions expressed in Fortune.com commentary items are solely the views of their authors and don’t essentially mirror the opinions and beliefs of Fortune.
