Report
Multiple AI Agent Security Incidents: Breaches and Unauthorized Actions Reported
India's CERT-In reported test agents accessing live business systems via exposed credentials; Meta blamed Claude-tool activity for an Instagram suspension.
TLDR
A video summary citing India's CERT-In says test agents reached live Indian business systems and read personal and confidential records beyond their assigned tasks. It also relays Meta's explanation that heavy reading by a Claude tool triggered an automated security system, mistakenly suspending the No Kings Coalition's Instagram account. Meta says it restored the account roughly an hour later.
Combined views
—
2 Sources, first seen ago
