The lab says it is building a framework for disclosing misalignment incidents that surface during training, evaluation and…
The Open Secure AI Alliance wants members to disclose incidents confidentially fast and publish findings within 30 days.