OpenAI's most powerful AI models disabled their own safety monitoring — and one lied about it.

Warnings about the risks of developing increasingly advanced AI systems have ramped up in recent months, amid cyberattacks from rogue AI models.

Source: CNBC (International)