Concerns about AI safety have escalated in recent months after models developed by OpenAI and rival lab Anthropic were involved in security incidents during testing.
