Anthropic said it disrupted attempts to misuse its Claude AI models for potential biological weapons research, weapons development and cyber espionage. In a new threat intelligence report, the company documented five cases involving researchers using Claude in ways that could support biological weapons development, including efforts to plan experiments involving avian influenza.
The company also reported that a suspected Russia-linked hacking group used AI extensively in cyber operations targeting Ukrainian government, military and diplomatic officials. Anthropic said the group used AI to automate activities including phishing, malware development and evading security detection. It also identified other criminal groups using Claude for weapons-related software and intelligence activities.
Anthropic further said it disrupted attacks involving seven China-based AI labs, including Alibaba, Moonshot and DeepSeek, which it accused of attempting to extract Claude’s capabilities. The company alleged that some operators conducted large-scale “distillation” efforts using Claude outputs to improve other AI models, while Moonshot and DeepSeek allegedly routed customer conversations through Claude for training purposes.
Pic courtesy: google/ images are subject to copyright
