The Malicious Use of Artificial Intelligence

A research report by Brundage and 25 other authors surveys potential security threats from malicious AI uses across digital, physical, and political domains, proposing strategies to forecast, prevent, and mitigate these threats. The authors provide four high-level recommendations for AI researchers and stakeholders while identifying promising areas for developing defensive measures and making attacks less viable. The report also examines the long-term dynamics between attackers and defenders in the context of AI-enabled threats.