Overview of AI Misuse Risks
Advanced AI tools can empower individuals of groups with harmful capabilities once limited to powerful states or specialists. Frontier models now offer expert-level reasoning and programming at low cost, lowering barriers for misuse. Malicious “uplift” occurs when AI amplifies an actor’s technical skills and strategic capacities.
Cyber Threats Enabled by AI
AI can rapidly scan digital targets, find vulnerabilities, and generate exploit code at machine scale. Simultaneous AI agents can overwhelm defenses and disrupt critical networks like utilities or financial systems. Automated cyberattacks blur attribution, making defensive prioritization difficult and costly.
Disinformation Tools
Advanced models create realistic deepfakes, audio, video, images, that mislead and destabilize public discourse. AI-generated content can fabricate crisis narrative or manipulate public sentiment at scale. AI chatbots can spread tailored falsehoods across platforms, enhancing propaganda reach.
Phishing and Social Engineering
AI crafts highly credible impersonations in email or messaging platforms. Automated personalization increases success rates for scams exploiting trust networks.
Biological and Chemical Threats
AI accelerates biological design, allowing novices to engineer dangerous pathogens. Tools like viral design platforms can enhance transmissibility or lethality without expert oversight. AI may also help optimize processes for toxin manufacturing or dispersion.
Targeting and Planning Enhancements
AI supports public data analysis to identify high-impact targets in infrastructure or institutions. Simulations enable malicious actors to refine strategies and anticipate defensive responses. Scenario planning tools reduce planning risks and increase coordination among low-skill groups.
Autonomous Platforms
Affordable AI perception systems may empower autonomous drones or robots. These platforms could locate and strike targets independently, evading detection. Distribution through open markets heightens the risk of unsupervised deployment.
Role of Global Powers in Risk Mitigation
The United States and China house most frontier AI models and massive computing power. Both nations share incentives to reduce malicious use that threatens stability and infrastructure.
Shared Safety Standards
Coordinating safety frameworks can limit high-risk capabilities across models and jurisdictions. Shared guidelines reduce “safety arbitrage,” where bad actors exploit weakest protections.
Threat Information Sharing
Exchanging data on attempted malicious prompts and behaviors defenses. Joint knowledge helps identify cross-border patterns invisible to single-nation monitoring.
Crisis Communication Channels
Dedicated AI “hotlines” facilitate rapid clarification during ambiguous attacks. Such channels can prevent escalation when attribution is obscured or weaponized.
Biosecurity Context
AI’s integration with biotechnology expands beneficial applications and dangerous misuse potential. Biological design tools without proper guardrails create step-wise instructions that non-experts can follow. Risks escalate when capabilities are paired with weak oversight or material access.
Towards Broader Cooperation
Mitigating AI misuse by non-state actors requires multi-layered global capacity building. Frameworks must balance preventing harm with allowing innovation and development. Expanded cooperation can support global capabilities, especially in regions with limited regulatory capacity.
Concluding Insight
AI misuse by non-state actors combines scalable automation with low cost and high impact. Cooperation among major AI powers is feasible, necessary, and beneficial for global stability. Addressing contextual risks is more effective than focusing solely on technological limits.
Source:
Chan, K., O’Hanlon, M. E., Haotian, Q., & Lefeng, Z. (2026, January 29). AI risks from non-state actors. Brookings. https://www.brookings.edu/articles/ai-risks-from-non-state-actors/