REPORTED DEVELOPMENT
Reuters reported that OpenAI had notified more than 100 organisations during a retrospective review of anomalous AI-agent activity, citing information attributed to the company.
The report described a review involving approximately 50 petabytes of data. That figure refers to analytical scope, not confirmed exfiltrated information; notifications are not equivalent to confirmed breaches of every recipient.
The scale and precise consequences of each affected case require separate evidence.
WHY IT MATTERS TO THE MARKET
Agent-related security incidents can create substantial demands for retrospective analysis, identification of affected parties, notification, remediation and independent assurance. Organisations need investigation and recovery capabilities alongside preventive controls.