Future Threats and Risks - I: Deepfakes, Autonomous Weapons, and AI Safety

Artificial intelligence (AI) is increasingly viewed as a general-purpose technology with thepotential to reshape production, communication, and governance. At the same time, it ischanging the threat landscape facing individuals, societies, and institutions. This chapteroutlines future AI threats and risks using a two-by-two framework that distinguishesepistemic versus kinetic harms and misuse versus systemic sources. Epistemic harmsdestabilize the information environment and the conditions for forming reliable beliefs, whilekinetic harms involve physical or cyber-physical damage. Misuse captures intentional,malicious deployment of AI; systemic risks arise from embedding AI in platforms, markets,and institutions even absent malign intent. The chapter then explores two case studies. First,deepfakes illustrate epistemic misuse by lowering the cost of credible deception and enablingthe contestation of authentic evidence, thereby weakening public trust. Second, autonomousweapons and AI-enabled targeting pipelines illustrate kinetic risk by distributing lethaldecision functions across human–machine ensembles, raising accountability challenges underinternational humanitarian law and increasing escalation pressures through speed andopacity. The chapter then reviews technical countermeasures, provenance standards, andemerging regulatory approaches, and explains why current responses lag behind rapidcapability growth. It concludes by assessing how agentic AI may amplify and recombine theserisks across domains.