Google reports first fully automated AI cyberattack by Russian-speaking hackers
Google said Monday that its threat intelligence team had identified Russian-speaking hackers who built AI agents capable of fully automating vulnerability discovery and exploitation with no human in the loop—what the company described as a first. The agents continuously compared code before and after patches, found flaws, and wrote exploits on their own. Technical documentation Google saw was in Russian. The campaign hit French websites in media, entertainment, and education.
Until now, even advanced models typically needed skilled operators to steer attacks or correct errors, while AI mainly sped up and focused human-led work. Google’s finding may therefore mark an inflection point in AI-enabled hacking, a debate that has intensified since Anthropic released its specialized model Mythos in April. Officials and operators of critical systems are already on edge after major flaws in PHP, a language said to underpin 70 percent of the web, and after supermarket-fridge hacks partly blamed on Iran. A Moscow-based firm known for large-scale information manipulation and interference is in the European Union’s sights; French Foreign Minister Jean-Noël Barrot said he came to provide colleagues with the proof. Experts treat Moscow as a leading suspect, though national authorities have not named a culprit. Google’s case appears tied to a recent wave of attacks on supermarket refrigerators across Europe.
Continue in the app — vote & join in ➔