Google tells that hackers use AI agents to make attacks from patches
Google tells that attackers use AI agent tools to turn a new patch into an attack in a short time.
On Monday, Google told that attackers use AI agent tools that do many attack steps at machine speed. Attackers also use AI to compare a patch with the code before it, and then attack the new vulnerability. Google gave no count of attacks and no test results, but four open source agent tools are in the report.
How it works
Some open source tools, made to test networks, are in live attacks at this time. Google tells that they do single steps and connect attack paths. Attackers also use AI to read code changes and release notes. Google gives the PaperCut NG/MF vulnerabilities of September 2026 as an example.
Who uses it
The users include groups that want money and suspected spy groups that Google connects to China. Jieyou sells phishing, which is fake messages that steal logins. It added a chatbot, and buyers with low skill change the fake messages in plain words.
What is not known
Google gives no number of victims and no speed figures in this report. Google tells that the basic methods stay the same. Only the speed and scale change.
Insight: Attackers shorten both ends of the time line
Google tells that AI agents turn a patch into an attack fast. An earlier Google report told that one group went from a first account to deep access in about four hours. In one attack the time was 15 minutes. This reads as a plan to cut the time to install a patch and the time to find the attacker.
Sources
Posted