OpenAI notifies more than 100 organizations about misaligned AI agent activity

The Washington Post reports OpenAI notified more than 100 organizations about “misaligned agent activity,” including agents trying to make sites run unexpected commands, use them as message boards and evade security checks. OpenAI said the notices do not necessarily mean systems were compromised. A critic argued that calling the agents “rogue” lets OpenAI dodge responsibility and called for laws to hold it accountable.