OpenAI’s agent findings show why enterprises need scoped credentials, per-request authorization, audit logs, and data controls alongside AI guardrails.

OpenAI has introduced a formal process for investigating and publicly reporting model behavior it considers unexpected or ...

OpenAI disclosed six misalignment reports describing models that used leaked API keys, uploaded data publicly and bypassed network limits.

One unreleased OpenAI model uploaded a file online without user permission to obtain a browser citation, while collaborating ...

OpenAI disclosed six cases where AI models ignored constraints, used exposed credentials, uploaded files, and attempted to ...

Learning from OpenAI Hacking Incidents! Security Measures and Practical Guides for AI System DevelopmentThe evolution of AI technology is remarkable, bringing innovation to our work and lives. However ...

The company published six training and evaluation cases and said the industry still lacks shared rules for saying when models go rogue.

A new OpenAI report reveals an AI model secretly gave itself a rogue new identity, along with five other cases of unexpected ...

On September 16, 2026, OpenAI released a "framework for reporting model misalignment."Announced at the same time were six instances of unexpected or concerning behaviors identified during training and ...