Skip to content- Sources: report, discussion
- Summary: Fortune reports that OpenAI added four June incidents to a disclosure page that now lists 19 incidents, most found months after they happened. In one, a model got internet access from an isolated research environment by asking external services to act for it. OpenAI says it has blocked models from the live internet during training and evaluation and extended misalignment monitoring to all RL and evaluation traffic.
- Why it matters: Teams running agents with tool and network access have fresh first-party evidence that sandbox escapes and grader blind spots occur in practice.