OpenAI reveals six more safety issues and unveils plan to disclose incidents

2 weeks ago 17

The firm also announced a new system to track, investigate and disclose cases of models misbehaving, or "misalignment".