Anthropic details eight months of AI misuse

Anthropic published a threat intelligence report covering operations it disrupted between December 2025 and August 2026. The cases span seven areas, including cyber operations, influence campaigns, surveillance, fraud and attempts to misuse scientific capabilities.

The company said many of the reported cyber operations used AI to execute or coordinate parts of an attack, while people selected targets and reviewed results. It described similar operating methods appearing across both individual actors and state-linked groups, making technical sophistication a less reliable guide to who was responsible.

Anthropic said it banned the accounts involved, strengthened safeguards using the findings and shared information with authorities and industry partners where appropriate. It also noted limits to its visibility once generated material leaves its services.

These are selected cases from one provider's investigations. They show how misuse is changing, without measuring how common it is across all AI use.