Google says attackers built AI framework turning patches into exploits automatically

For the CISO of a regional US bank

Google's Threat Intelligence Group disclosed on Monday that it found an AI framework, identified as exploit_agent, that attackers built to turn software patches into exploits automatically, according to VentureBeat's report on Google's findings.1

Google says it is the first time it has identified attackers using LLM frameworks to fully automate vulnerability discovery and exploitation "without a human in the loop," according to VentureBeat.1 The framework is set up to continuously compare code before and after patches, identify potential vulnerabilities, draft exploits and run them without human intervention, according to Google.12

Researchers found exploit_agent on an exposed staging server alongside stolen victim data and post-compromise scripts, but the report does not establish that the framework itself was responsible for those compromises.12

Google did not identify a specific vulnerability exploited by the framework, provide a victim count or report a measured patch-to-exploit time.1

The operator's configuration file contains Russian-language comments that describe exploit_agent as a pilot and tell the operator to enter LLM API keys directly into the file, according to Google; Google does not attribute the framework to a particular group or country.12

Victim telemetry on the server suggests the operator was interested in French media, entertainment and education websites, according to Google.12

Why it matters

If attackers can turn a published patch into a working exploit without a person doing the analysis, the time between a vendor's patch release and attempted exploitation of unpatched systems may shrink. Banks and fintechs should check how fast they patch internet-facing systems and whether they monitor for exploitation attempts, not only for new vulnerabilities.

From the brief

Our take. Every patch tells two readers what was broken: your team, which queues the fix, and a script that reads it overnight. A critical-patch SLA measured in weeks assumes the attacker also needs weeks.

Action. Vulnerability management: Shorten critical-patch SLAs for internet-facing systems to days and add exploitation-attempt alerting on systems awaiting patches.

Sources

  1. ReportGoogle says attackers built AI framework turning patches into exploits automatically Venturebeat
  2. Also reported byGoogle says an AI framework turned security patches into exploits with no human in the loop Siliconangle