Anthropic just released a terrifying BOMBSHELL threat intelligence report. - The report describes cyberattacks in which Russian groups used AI agents to rebuild malware until it evaded antivirus detection. The group hijacked DNS on hotel Wi-Fi, stole drone-vision SDKs and mass-exported WhatsApp data. - Anthropic claims Iran-linked operations used Claude to target U.S. Navy bases, compile dossiers on Americans and run influence campaigns across U.S. social media. They even uploaded plans for the Ayatollah’s funeral. - Chinese-linked AI agents reportedly identified 12+ potential zero-days, decompiled 1.8M Android apps, stole 1TB+ of identity data, breached an airline, and extracted 2,100 Azure tokens from 40+ companies in 34 hours. - In one North African government breach, attackers allegedly pulled more than 300,000 identity documents and records on over half a million companies. - Anthropic says some captured conversations involved sensitive material, including Russian defense documents, Chinese police traffic and requests linked to the Chinese military. - In one case, Anthropic says a weapons group used Claude to help develop and test a guided rocket. After the test failed, the group returned to the model to diagnose what went wrong. Then comes the model-distillation allegation: • Alibaba’s Qwen allegedly generated more than 151 million Claude exchanges between May and July, peaking at nearly 3 million per day. • Moonshot allegedly routed nearly 300,000 requests to Claude through roughly 5,380 fake accounts in about 10 days, presenting Claude as Kimi to users. • DeepSeek allegedly used similar routing, including Claude Code and OpenCode traffic. The report’s warnings highlight how AI is evolving beyond a tool for attackers into the infrastructure powering the next generation of cyberwarfare.