Google’s Threat Intelligence Group has captured a criminal hacking crew using an AI-based zero-day vulnerability in the wild for the first time, neutralizing a planned mass attack before it could be carried out.
This finding falls within a broader report showing that attackers are now weaving large linguistic patterns into every stage of the intrusion. Defenders race to deploy their own AI Hunters in the same battle.
How the Zero-Day AI vulnerability worked
Malicious code written in Python has been bypassed Two-factor authentication (2FA) A popular open source system administration tool. Google didn’t do that His name The affected seller.
Follow us on XTo get the latest news as it happens
Several references point to the author of a major linguistic model. The script contains tutorial-style documentation and a fabricated score for the Common Vulnerability Scoring System (CVSS), a metric that no human researcher would have invented.
Google He said His Gemini model was not used. John Hultquist, GTIG’s senior analyst, warned that precision AI-powered intrusions may already be underway undetected.
“Each new generation of models will reduce the need for expert-developed belts, but they will almost certainly exist. We have to realize the limits of our visibility into the backend of spies and criminals. The signs will not be clear. The race has already begun.” He said.
The defenders retreat
The same report cited Russia-linked malware families PROMPTFLUX and PROMPTSPY, an Android backdoor that sends a communication to Gemini in real-time to plan its next action.
State-linked Chinese and North Korean operations train special models on a dataset of 85,000 vulnerabilities.
Google countered this with Big Sleep, an AI agent that hunts down a day before attackers can find them, and CodeMender, an automated correction system. Big Sleep has already closed a loophole that hackers were preparing to weaponize.
Why should cryptocurrencies be monitored?
The gap between attack and defense is getting sharper. Binance research recently found this AI agents exploit smart contracts Twice as well as detecting threats.
Previous reports have indicated how Google AI tools do this Helping scammers drain walletsand a new bug in Chrome recently Exposed private keys.
Against this background, exchanges take place Deploy their AI shieldsbut the bar keeps going up.
With both sides now fielding independent agents, the next Zero Day could emerge from a machine on either bench.
this post Google Discovers First AI Vulnerability in Zero-Day: A Warning Shot for Cryptocurrency Security? appeared first on BeInCrypto.
