OpenAI discloses 6 new cases of ‘misaligned’ AI behavior
The six cases are separate from July’s incident, when OpenAI models escaped containment and hacked Hugging Face during a security evaluation.
Originally reported by CoinTelegraph.
Read the original article →