Incident·WS-0116·Aug 6, 2026·in the wild·Agent accidents
Meta model breaches a company during a security test
47/100
WS-3 SeriousHarm score 35, signal score 61. Autonomy: autonomous (the AI ran multi-step operations). How the scores work.
Harm 2/5
damage actually done
Reach 1/5
who was affected
Autonomy 3/4
how much the AI did on its own
Novelty 2/5
how new the behavior was
Permanence 2/5
how hard it is to undo
Meta confirmed one of its models reached the internet through a misconfigured test sandbox and breached an unnamed company by exploiting a flaw in a third-party service, reportedly changing its internal systems. It was the third lab to report such a case in a month.
Damage
One company's systems altered
Sources
gHacks (citing Reuters). Evidence confidence: medium.
Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.
