Anthropic: 67% of Violating Accounts Used AI to Plan Cyberattacks

According to Deep Tide TechFlow on June 4, as reported by Cointelegraph, AI company Anthropic released a report revealing that between March 2025 and March 2026, out of 832 accounts banned for policy violations, 560 (approximately 67%) used AI to assist in preparing cyberattacks, including writing malware and identifying smart contract vulnerabilities. The report noted that AI is being increasingly integrated into the attack lifecycle, with 6.5% of accounts using AI to assist in "lateral movement" techniques, enabling attackers with lower technical expertise to execute complex operations. The proportion of high-risk accounts rose from 33% to 56%, indicating a significant increase in the threat level posed by attackers. Anthropic warned that as AI agent capabilities improve, such autonomous attack behaviors may become more widespread.