Skip to content

Frontier AI Risk Management Framework

July 2025

Shanghai AI Laboratory and Concordia AI are proud to introduce the Frontier AI Risk Management Framework v1.0 (the “Framework”). We propose a robust set of protocols designed to empower general-purpose AI developers with comprehensive guidelines for proactively identifying, assessing, mitigating, and governing a set of severe AI risks that pose threats to public safety and national security, thereby safeguarding individuals and society. 

This framework serves as a guideline for general-purpose AI model developers to manage the potential severe risks from their general-purpose AI models. The framework encompasses six interconnected stages: Risk Identification categorises four major risk types (misuse, loss of control, accident, and systemic risks); Risk Thresholds establish a set of unacceptable outcomes (red lines) and early warning indicators for escalating safety and security measures (yellow lines); Risk Analysis recommends conducting risk analysis throughout the entire AI development lifecycle to determine whether the AI has crossed the yellow lines; Risk Evaluation classifies models into green, yellow, or red zones based on risk levels; Risk Mitigation outlines defense-in-depth approaches spanning safety training, deployment measures, and model security measures; and Risk Governance provides oversight mechanisms including internal governance, transparency and social oversight, emergency controls, and policy updates.

Recognising AI safety as a global public good, we call for widespread adoption of compatible frameworks across the industry. This collaborative approach is essential to ensure that transformative AI technologies benefit humanity while avoiding catastrophic risks, fostering a global “race to the top” in AI safety standards.

Alongside the Framework, Shanghai AI Laboratory released a technical report (Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report) that evaluated several frontier models across seven key areas: (1) cyber offense, (2) biological and chemical risks, (3) persuasion and manipulation, (4) strategic deception and scheming, (5) uncontrolled autonomous AI R&D, (6) self-replication, and (7) collusion. Concordia AI led the biological and chemical risk evaluation of the technical report.

[English Version] [Chinese Version]

Authors: Shanghai AI Laboratory and Concordia AI
Back To Top