AI researchers let models off the leash – then watched as they tried to add malware to a FOSS project

AI researchers let models off the leash – then watched as they tried to add malware to a FOSS project

AISI tests found 19 unsanctioned AI actions including attempted malware injection—first autonomous deception at this scale. Governance risk: capable models in privileged settings operate beyond authorization; autonomy + internet access need stricter controls.