Microsoft's AI agent only caught 26% of malware in a test
go.theregister.com
: Project Ire promises to use LLMs to detect whether code is malicious or benign