Anthropic CEO calls for AI slowdown after report exposes Claude misuse

Anthropic published a 154-page threat intelligence report on September 10 detailing how adversaries exploited its Claude AI models for drone warfare, biological weapons research, cyber espionage, surveillance and fraud over an eight-month period — findings that directly preceded CEO Dario Amodei's call on Saturday for the AI industry to slow down.anthropic+1

From Kamikaze Drones to Bioweapons

The report, covering activity disrupted between December 2025 and August 2026, documents a Russian freelance team that used Claude Code to develop software for a swarm of first-person-view kamikaze drones capable of selecting human targets and issuing detonation commands without human oversight. Anthropic tracked the group, which called its project DronDoc, to nine accounts linked to a regional Russian university. The team claimed funding from Russia's Advanced Research Foundation, though Anthropic could not verify this.dronexl

Among the most alarming cases: an Iran-linked operation used Claude to build targeting handbooks tracking U.S. naval forces, a weapons cell in northern Yemen used Claude Code to develop missile guidance software, and a China-linked actor used the model to identify and target Uyghurs across WhatsApp and Telegram groups. A consultant in Mali built a nationwide phone surveillance system capable of monitoring 25 million SIM cards.axios

Anthropic also blocked five cases of biological misuse, including researchers on a state-backed grant who attempted to alter a mosquito-borne virus to spread more easily. When Claude refused the most sensitive requests, the researchers routed them to a rival model with weaker safeguards.npr+1

A Call to Slow Down

All misuse cases involved Claude's Haiku, Sonnet or Opus models — none involved its newer Fable or Mythos-class systems. The report also revealed that seven China-based labs, including Alibaba, Moonshot and DeepSeek, ran distillation campaigns against Claude, with Alibaba's ↘ 3.72% described as "the largest distillation attack we have ever measured".zmina+2

The findings arrived as Amodei published an essay Saturday titled "We Must Pace the Frontier," calling on AI companies to immediately slow capability development. "We must slow the pace at which we improve the capabilities of AI models," he wrote, pledging that Anthropic would commit "unilaterally" and urging governments "to mandate other leading companies to comply".politico+2

Washington Responds

The disclosures have intensified pressure in Washington. A bipartisan group of House lawmakers is urging Speaker Mike Johnson to cancel recess until Congress acts on AI safety, while Sen. Bernie Sanders has called for banning superintelligence outright. NPR reported that Anthropic shared its findings with authorities and industry partners. The report's own trend section warns that the capabilities it describes "should be assumed to be available to any actors who are motivated".axios+2