Russian espionage, dissident monitoring and AI cyberattacks: What Anthropic’s report says

In a recent report titled 'Detecting and Countering Misuse of AI: September 2026,' Anthropic detailed how its AI models were misused in various malicious activities between December 2025 and August 2026. The report, the company's fourth, revealed how attackers have been leveraging AI for cyberattacks, espionage, surveillance, and fraud.

The company's Threat Intelligence Team identified these cases and disrupted the activities, using findings to enhance safeguards. The report covered seven harm areas, including cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development, and distillation. While none of the attacks utilized Anthropic's most advanced models like Claude Fable or Mythos, except for an instance of illicit distillation, they showed how AI is evolving to orchestrate cyber operations directly, reducing the need for human involvement.

AI is also lowering the barriers to launching sophisticated attacks, as it can assist in reconnaissance, tool development, data processing, and exfiltration. This could allow less technically skilled individuals to carry out attacks that previously required significant expertise. The report also highlighted a threat actor known as GTG-20006, believed to be linked to Russia's Midnight Blizzard, which targeted various military, government, and defense organizations using AI across multiple stages of their operations.

The use of AI significantly accelerated these attacks, making them faster, cheaper, and more repeatable.