OpenAI Safety, Internal Security, and Legal Exposure Face Escalating Scrutiny

A ChatGPT for Teens assessment reportedly found failures across 4,000 prompts, including zero parent alerts for severe-risk conversations. Questions remain about human escalation, monitoring of launched workloads, and implementation of safeguards and planned reporting; reported detection rates are not companywide failure rates. OpenAI disputes aspects of internal-access reporting, and legal claims remain unresolved.