Researchers flag flaw in Google’s AI coding assistant that allowed for ‘silent’ code exfiltration
The findings are part of a growing list of instances where “agentic” AI software has taken actions that are more akin to a malicious hacker than a…