ResearchUpdated 2 min read1 source
Google found an attack framework set up to turn patches into exploits automatically
Google Threat Intelligence Group found, on an exposed server, a framework set up to compare code before and after a patch, draft exploits and run them.
At a glance
- Date
- Topic
- Research
- Reading time
- 2 min
- Sources
- 1
This entry is compiled automatically with AI and is not read by a person before it is published. Its sources are below. If you see an error, write to us.

In a report dated 16 November 2026, the Google Threat Intelligence Group (GTIG) said that in the past quarter it saw, for the first time, attackers using LLM frameworks to automate finding and exploiting vulnerabilities from start to finish, with no person involved.
On an exposed staging server GTIG found a multi-agent framework named exploit_agent. According to the report it is built to work through popular web applications continuously: it compares the code from before and after a patch announcement, picks possible flaws out of the differences, writes draft exploits for them and runs the code without anyone approving each step. The server also held data taken from victims and scripts for use after a compromise.
GTIG says frameworks of this kind can significantly compress attack timelines and shrink the patch window. As one example the report includes a figure of the framework examining the access-control code in version 2.19.25 of the Spectra (Ultimate Addons for Gutenberg) WordPress plugin. It does not say whether an exploit worked in that case.
GTIG says the victim data on the server suggests the operator was specifically interested in targeting French websites associated with the media and entertainment industries, as well as the education sector.
What it means for you
The studio’s noteThis is our own reading, not the report’s advice: if you run a site on ready-made web software such as WordPress, applying security updates within the week they are published, rather than in an end-of-month batch, looks sensible. It is a way to reduce risk, not a guarantee.