AI Will Not Save Cybersecurity

In September 2025, Anthropic's security team detected something unprecedented. An AI system was being used not merely as an advisor to human hackers, but as the primary operator of an espionage campaign. At its peak, the AI made thousands of requests per second, probing systems, adjusting tactics, and exploiting vulnerabilities at a pace no human could match. According to Anthropic's analysis, the threat actor was able to use AI to perform 80 to 90 per cent of the entire campaign, with human int