01 / AI · Analysis
Google reports AI attacks that analyse and exploit patches
The significance of reported automation from patch analysis to attack execution lies in how much work proceeds without human intervention.
Google Threat Intelligence Group (GTIG) and Mandiant reported automated patch analysis and attacks in an AI threat tracking report, according to Korean technology publication TechM; independent replication has not been confirmed.
TechM reported an analysis suggesting that AI use in attacks is shortening the time companies have to apply security patches. The central issue is the process linking an attacker’s analysis of a patch to the execution of an attack.
TechM described the cases as different from earlier investigations involving AI-assisted vulnerability discovery or automation of individual attack tasks. The distinction turns on the range of work that continues without human intervention.
From patch analysis to attack execution
According to TechM, the reported system uses AI to analyse security patches, generate attack code and execute it. This allows the process to proceed without hackers finding vulnerabilities or writing the code themselves.
TechM reported that Google Cloud, Google’s cloud computing business, attributed the findings to GTIG and Mandiant, its threat intelligence and cybersecurity teams. The scope described by the company includes vulnerability discovery and attack automation without human intervention.
Assessing the scope of autonomy
The confirmation of the cases reported by TechM comes from Google Cloud. A company’s account of activity it has observed needs to be distinguished from results independently reproduced by researchers.
TechM reported that earlier investigations had observed automation of parts of the attack process. The significance of the latest autonomy claim therefore rests on which tasks previously performed by people now run together automatically, from analysing a patch through to executing an attack.