Meta has launched its Frontier AI Framework, a policy document outlining criteria for the release of advanced AI systems that may pose risks.

The framework defines two categories: “high risk” and “critical risk,” which Meta considers too dangerous for public or external use.

Both categories encompass AI technologies that could enable cybersecurity breaches, as well as chemical or biological attacks.

Critical-risk systems could cause “catastrophic” consequences that cannot be controlled, while high-risk systems may facilitate harmful actions but with less certainty.

The framework aims to balance transparency with caution in the development of AI technology.

Recent Meta developments