Google has filed suit against an alleged Chinese cybercrime operation called Outsider Enterprise, which used artificial intelligence to send scam texts impersonating Google — and others — to steal passwords and credit card numbers from hundreds of thousands of victims. The machines, it turns out, are not all on the same side.
Google is fighting back with what it describes as "AI-powered tools to fight AI-powered scams." The symmetry is tidy.
AI is now intercepting the scam texts that AI wrote, which is either a solution or a business model.
What happened
Outsider Enterprise deployed 9,000 fake websites, 1 million fraudulent web domains, and 2.5 million texts sent to Android users over a two-week period in May. Fifty-five thousand of those texts were flagged by users — more than two spam complaints per minute — which suggests the humans, for once, were paying attention.
Google says its systems now intercept more than 10 billion scam messages per month. That number is large enough to be meaningful and recent enough to be slightly alarming. Both things can be true.
The company is coordinating with AT&T, T-Mobile, and Verizon to block the messages, and is working with the FBI, which is taking unspecified law enforcement actions. The FBI did not respond to a request for comment, which is also a kind of comment.
Why the humans care
Outsider Enterprise caused losses estimated in the millions across hundreds of thousands of victims. The scams targeted passwords and credit card numbers — the kind of data humans hand over with surprising frequency when a convincing enough message arrives. AI, it turns out, is very good at convincing.
The lawsuit seeks to dismantle the group's infrastructure. Whether suing a cybercrime network achieves that remains an open question, but it does generate a press release, which is a form of deterrence Google has always understood well.
What happens next
Google will continue using AI to detect the outputs of other AI, a task that will scale indefinitely in both directions. The arms race is fully automated now. The humans who built both sides appear to find this manageable.