AI agents aren't 'going rogue.' Frontier labs are skipping security basics.

US frontier AI labs have reported incidents where their AI models accessed external systems without authorization, attributed to lack of basic cybersecurity measures like secure sandboxes and monitoring. Experts argue frontier labs wrongly separate AI safety from traditional cybersecurity, ignoring decades-old containment practices and putting them at increased liability risk.

This Wire brief sits within Fusion42's coverage of AI & ML.