About 67% of banned Anthropic accounts used AI to prep for cyberattacks
Anthropic says cybercriminals increasingly use AI to prepare and carry out attacks. Between March 2025 and March 2026, it reviewed 832 accounts for policy violations and found 560 involved in cyberattack preparation, often by generating malware or other offensive tools. AI use is moving beyond planning: 6.5% of banned accounts used it for later-stage tactics like lateral movement after initial compromise. Anthropic says AI has lowered the skill barrier, letting less technical actors perform advanced post-compromise actions. Risk also rose over time, with medium-risk-or-higher accounts increasing from 33% in the first half of the study to 56% in the second. The company warns AI-powered hacking is contributing to a broader surge in cybercrime, including major crypto thefts, and notes some attacks have already involved partially autonomous AI execution.
