An internal probe into security failures led to the dismissal of several alignment staff on Oct. 1. The employees allegedly shared confidential company information with a third party AI safety organization and mishandled sensitive data outside established corporate procedures, violating company policies, the Wall Street Journal reported.

The terminations follow a month of model pauses and the scrapped launch of GPT 6.1 Astra, as safety and reliability requirements become primary constraints for more capable AI agents. The move arrives amid intensified industry-wide scrutiny of AI security risks and the internal governance of alignment research teams.