Tech bro's "anti-fraud" AI actually just regular old fraud

The head of an AI advertising company that promised "a world free of fraud" just got sentenced for fraud.

As reporter in Ars Technica, Paul Roberts, CEO of Kubient, sold investors on magical AI technology that could supposedly catch advertising fraud in milliseconds. His company's IPO materials bragged about "machine learning powered pre-bid ad fraud prevention technology" and a "self-learning neural network always getting smarter." — exactly the kind of Theranos-style word salad that separates venture capitalists from their money.

Roberts, 48, wrote that Kubient was "identifying and preventing approximately 300% more digital ad fraud" than competitors. Spoiler alert: It wasn't. Instead of using it to catch bad guys, Roberts had his employees generate completely fake reports based on imaginary data to fool auditors and investors.

The scheme involved a bogus $1.3 million transaction with another company where they basically played financial hot potato, passing the same money back and forth while pretending to provide services that never actually happened. This phony revenue made up 94% of Kubient's reported income during their IPO.

But you can't neural network your way out of federal charges. The SEC, postal inspectors, and federal prosecutors came knocking, and Roberts is now headed to prison.

"Paul Roberts cooked the books," said Acting U.S. Attorney Matthew Podolsky, probably resisting the urge to add "and now he's cooked his own goose." "He lied to investors and auditors about his company's revenue and about his company's premier product."

Roberts company is in bankruptcy liquidation, naturally, though they've merged with something called "Adomni" — presumably because "Web3MetaverseAI.io" was taken.

Previously:
• Crypto bro Do Kwon dragged to US after $40 billion collapse and global manhunt
• How many Wells Fargo employees were fired for NOT committing fraud?
• Britain's AI-based benefit fraud detector unfairly targets minorities