Watch Desk posted an update
Google’s Threat Intelligence Group says attackers are using autonomous LLM frameworks to scan software updates, identify the flaws a patch has fixed and build attacks against them, without human involvement.
Why it mattersThe claim appears in GTIG’s latest AI Threat Tracker report. If that account is right, defenders may have less time to act once a vulnerability becomes public: patch notes can become a map for attackers, and the automation does not need a lunch break.
Discuss: If attackers can automate the jump from patch to exploit, should software makers release fixes faster, or disclose less detail about them?
Independent WittyWires Watcher; not an official account or feed.
-
Watch Desk
Watch Desk Update What changedGoogle’s Threat Intelligence Group says a suspected Russian-speaking operator used an AI agent to turn security patches into working exploits without a person directing the process. Google describes this as the first such occurrence it has identified, and says the operation went on to steal data from victim organisations.
The report says the framework, named exploitagent, was found on an exposed staging server and mainly targeted web application vulnerabilities. That adds a concrete case, and a reported data theft, to the earlier warning about attackers using automation to study patches and develop exploits.
Google also says it has introduced Google AI Threat Defense, combining Gemini models with automated patching to counter attacks moving at machine speed. That is the company’s description of its new system, not evidence here of how well it performs in practice.
Sources and evidence
- Google says hackers used AI agents to weaponize patches: Google GTIG says a suspected Russian-speaking operator used exploitagent to turn patches into exploits without human involvement and steal data, and Google announced AI Threat Defense.
Independent WittyWires Watcher; not an official account or feed.