AI agents found a zero-day on their own
Attackers used the open-source tool Cairn
Security
Google Threat Intelligence Group · first with no human in the loop
Written with AI assistance, reviewed by the ACN desk.
“In the past quarter, GTIG identified threat actors using LLM frameworks to fully automate vulnerability discovery and exploitation without a human in the loop for the first time.”
Google has reported the first attacks it has seen run end to end by AI. Google Threat Intelligence Group’s report says threat actors used large language model frameworks to automate vulnerability discovery and exploitation.
The group says it saw this in the past quarter “without a human in the loop for the first time.” One tool, exploit_agent, compares code before and after patch announcements. It then drafts exploits and runs them without human intervention.
Google says this is shrinking the patch window for organizations. SiliconANGLE notes that the previous quarterly edition of the report had not observed fully autonomous pipelines.
BleepingComputer reports that Google’s researchers expect sophisticated, fully autonomous agentic attacks to emerge in the next six months. The report was published on November 16.
“In the past quarter, GTIG identified threat actors using LLM frameworks to fully automate vulnerability discovery and exploitation without a human in the loop for the first time.”
Sponsor
Spot an error? Email corrections@aicodingnews.si. Fixes are logged on the corrections page.
Sponsor
Sponsor
Attackers used the open-source tool Cairn
A synthetic internet for prompt-injection tests
F5 patches HTTP/2 code-execution bugs · found by five teams
CVE-2026-114938 · Oracle has released a patch
Sponsor
Sponsor