Anthropic Details Disrupted Claude Misuse Across Seven Harm Areas

Anthropic's September 2026 report reveals how its Threat Intelligence team disrupted misuse of its Claude AI models from Dec 2025 to Aug 2026 across seven harm areas including cyberattacks and illicit model distillation. The report details specific threat actors, including those linked to Midnight Blizzard and ShinyHunters, targeting government and commercial entities with AI-enhanced operations. Anthropic used these findings to improve safeguards and shared intelligence with authorities and the