Skip to content
AI news · Privacy & rules

Attackers now have an AI that turns security patches into break-ins, which makes 'I'll update it later' an expensive habit

By , Founder of Adaptd ·

Google's threat team found an AI framework that reads new security patches, writes an attack for the flaw and runs it with no person involved.

The short version

  • Google's Threat Intelligence Group found an AI framework that compares software before and after a patch, then writes and runs an attack.
  • Google says it is the first fully automated attack pipeline it has seen, in a report published on 17 November 2026.
  • Any business with a website, plugins or internet-facing servers should cut the time between a patch coming out and being installed.

A software update used to give you a head start. When a fix came out, attackers still had to work out what it fixed. Google says AI is now doing that homework for them.

What did Google find?

Google's Threat Intelligence Group found an attack framework called exploit_agent that turns security patches into working attacks without a person involved. It was sitting on an exposed staging server, connected to several frontier and locally run AI models.

The framework keeps comparing popular web software before and after each patch announcement. When the difference points to a likely flaw, it drafts an exploit and runs it. One screenshot in the report shows it checking the access controls in a popular WordPress plugin.

How fast are attackers moving now?

Attackers can now have a working exploit within a day of a flaw being made public. In late August, researchers and the security company Huntress used AI to build working exploits for two PaperCut print server flaws within a day, and GreyNoise reported an attack campaign four days after disclosure.

Google also says published flaws and exploits nearly doubled between January and August 2026. The patch notes have become a treasure map, and the robots are reading them faster than we are.

What should your business do about it?

Your business should treat security updates for anything facing the internet as a same-week job. That means the website and its plugins, the firewall, remote access tools and anything else people can reach from outside.

My advice: ask whoever looks after your website how long a security patch usually waits before it's installed, and delete the plugins nobody uses. Think you're too small to be a target? This framework doesn't pick victims by size. It goes after popular software, so if you run it, you're on the list.

Attackers have built a robot that never knocks off. The least we can do is click Update.

Wondering what this means for your business?

That's exactly what the discovery call is for.