Meta has unveiled a new AI risk management framework to address concerns around the potential misuse of artificial intelligence. The Frontier AI Framework categorizes AI risks into two categories: high-risk and critical-risk, with the latter being severe enough to cause catastrophic consequences.
According to the framework, high-risk AI models may aid in cyber or biological attacks, while critical-risk AI can cause severe harm. To mitigate these risks, Meta will pause the development of critical-risk AI and restrict access to high-risk AI until additional safety measures are implemented.
The framework outlines a structured review process, with senior decision-makers overseeing final risk classifications. Meta will also rely on assessments from internal and external researchers to determine AI system risk levels. No single test can fully measure risk, making expert evaluation a key factor in decision-making.
This move is part of Meta's efforts to prioritize AI safety and responsibility. The company has pursued an open AI development model, allowing broader access to its Llama AI models. However, concerns have emerged regarding potential misuse, prompting the development of the Frontier AI Framework.
Meta's framework is similar to the AI Risk Management Framework (AI RMF) developed by the National Institute of Standards and Technology (NIST), which provides a structured approach to managing AI risks.
