Anthropic details AI-driven misuse cases in new report

Anthropic published a report summarizing 117 misuse cases detected across its Claude models, including cyberattacks, influence operations, and surveillance workflows. The company described AI agents automating reconnaissance, exploitation, data theft, propaganda production, surveillance, and research, while humans selected targets, set goals, and reviewed important outputs.

Attackers used AI to industrialize credential theft, cloud compromise, phishing, vulnerability research, and the extraction of sensitive data from downstream organizations. Influence operations relied on persistent agent memory, fake news sites, fabricated journalists, synthetic personas, political profiling, and large-scale multilingual content, though high-volume content often yielded little genuine engagement.

Surveillance and repression cases involved automated dossiers, biometric and communications analysis, transnational targeting, coercive recruitment, and systems that continued operating locally after model access was revoked. The report also documented dual-use risks in biological and weapons contexts, noting AI support for advanced scientific and military work, but did not establish completed biological weapons or operational battlefield deployment.