Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report

2025-07-22

AI Summary

This study systematically identifies and quantifies potential harms posed by state-of-the-art AI models across seven emerging risk domains: cyberattacks, biochemical threats, persuasive manipulation, uncontrolled autonomous R&D, strategic deception, self-replication, and collusion. Method: We propose a dynamic risk assessment framework integrating the Environment–Threat–Capability (E-T-C) analytical model with the AI-$45^circ$ Law, enabling the operationalization of abstract risks—particularly strategic deception and self-replication—into measurable, empirically grounded metrics. A novel three-tier threshold system (red/yellow/green) is introduced for risk classification. Contribution/Results: Empirical evaluation shows that current mainstream models remain below the red threshold, with most residing in the green or yellow zones. Persuasive manipulation consistently reaches the yellow threshold; certain advanced reasoning models approach the yellow threshold in strategic deception and self-replication dimensions. The framework delivers a practical, scalable risk stratification paradigm and actionable response guidelines for AI safety governance.

Technology Category

• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness
• Natural Language Processing: Safety and Robustness
• Humans and AI: Learning Human Values and Preferences

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

To understand and identify the unprecedented risks posed by rapidly advancing artificial intelligence (AI) models, this report presents a comprehensive assessment of their frontier risks. Drawing on the E-T-C analysis (deployment environment, threat source, enabling capability) from the Frontier AI Risk Management Framework (v1.0) (SafeWork-F1-Framework), we identify critical risks in seven areas: cyber offense, biological and chemical risks, persuasion and manipulation, uncontrolled autonomous AI R&D, strategic deception and scheming, self-replication, and collusion. Guided by the "AI-$45^circ$ Law," we evaluate these risks using "red lines" (intolerable thresholds) and "yellow lines" (early warning indicators) to define risk zones: green (manageable risk for routine deployment and continuous monitoring), yellow (requiring strengthened mitigations and controlled deployment), and red (necessitating suspension of development and/or deployment). Experimental results show that all recent frontier AI models reside in green and yellow zones, without crossing red lines. Specifically, no evaluated models cross the yellow line for cyber offense or uncontrolled AI R&D risks. For self-replication, and strategic deception and scheming, most models remain in the green zone, except for certain reasoning models in the yellow zone. In persuasion and manipulation, most models are in the yellow zone due to their effective influence on humans. For biological and chemical risks, we are unable to rule out the possibility of most models residing in the yellow zone, although detailed threat modeling and in-depth assessment are required to make further claims. This work reflects our current understanding of AI frontier risks and urges collective action to mitigate these challenges.

Problem

• Identify risks from advanced AI models in seven areas
• Evaluate risks using red and yellow lines for thresholds
• Assess current AI models' risk zones for safety

Innovation

• Uses E-T-C analysis for risk assessment
• Implements AI-45 Law with risk zones
• Defines red and yellow lines thresholds