AI Espionage Unveiled: Anthropic Disrupts Global Threats

Anthropic claims it has thwarted numerous malicious uses of its Claude models in the past eight months, including a Russia-affiliated cyberattack and unauthorized adaptations by Chinese AI firms. The report highlights how cybercriminals and state-sponsored hackers are increasingly utilizing AI to execute sophisticated cyberattacks and software development for weapons.

Anthropic announced on Thursday that it has disrupted various malicious activities involving its Claude models over the past eight months, including a suspected cyber espionage campaign thought to be linked to Russia and attempts by Chinese AI firms to copy Claude's capabilities.

The company revealed in its latest Threat Intelligence report that AI is increasingly being used by cybercriminals and state-backed hackers to execute complex cyberattacks. Rather than maintaining simple interactions with AI, these entities are employing multi-agent frameworks to carry out extensive tasks.

Notable disruptions included efforts from China-based labs such as Alibaba, Moonshot, and DeepSeek attempting to extract Claude's capabilities. Anthropic also linked Russia-based hackers, like the Midnight Blizzard group, to various attacks against Ukrainian interests. The report identifies new categories of threats involving the development of conventional weapons using AI, emphasizing the growing danger posed by these sophisticated methods.