Anthropic has lifted the lid on a disturbing wave of misuse, revealing that criminals, state‑backed groups and even scientists have tried to bend its powerful Claude AI toward designing missiles, conjuring deadly pathogens and spying on dissidents. The stark warning comes just 48 hours after a former researcher quit, warning that the firm’s race toward superintelligence could trigger human extinction by 2030.
Inside Anthropic’s Threat Intelligence Report
Published on Thursday 10 September 2026, the 154‑page threat intelligence document details five case studies where researchers sidestepped Anthropic’s safety guards to pursue biological work. The company stressed that these examples are not everyday misuse but the most notable and novel threats it has observed to date. “We’re publishing this work because we believe we have a responsibility to disclose malicious misuse of our services,” Anthropic wrote.
Biological Research Red Flags
Among the highlighted cases, a scientist employed Claude to draft a state‑sponsored grant application aimed at studying the chikungunya virus – a mosquito‑borne illness akin to dengue and malaria that can leave victims with months of severe pain, fever and debilitating symptoms. Anthropic told the New York Times that the project’s planned location at a military research institute raised particular alarm, noting that such knowledge could equally serve vaccine development or weaponisation. The firm said it banned the accounts involved but declined to name the institutions or countries, citing uncertainty about the researchers’ true intent.
Cyber Espionage, Surveillance and Propaganda Tactics
The report goes far beyond the lab, cataloguing dozens of other misuses. Russian actors were caught using Claude for espionage and “smash‑and‑grab” cyber‑hacks. A China‑based operation targeted Uyghurs in Syria, while another surveilled internal dissidents. Propaganda campaigns leveraged the model in Russia, Malaysia, Iran and Bangladesh. In Yemen, China and Russia, Claude helped write software for conventional weapons – ranging from firearms and missiles to armed drones and bombs.
The Fallout from Jacob Coxon’s Resignation
The report’s release follows a dramatic departure just two days earlier. Jacob Coxon, a former Anthropic employee, resigned publicly, claiming the company was not acting responsibly in its AI development. He warned that Anthropic and its rival OpenAI were “racing straight to self‑improving superintelligence” that could cause human extinction by 2030. Current staff echoed his concerns in public posts, intensifying the debate over AI safety.
Why it Matters
Anthropic’s disclosures underscore a pressing reality: the same generative power that fuels breakthroughs in medicine and creativity can be hijacked to craft biological threats, cyber weapons and disinformation at speed. As models grow more capable, the potential for catastrophic misuse expands unless developers, governments and civil society unite to build robust safeguards. The stakes are not speculative; they are already playing out in labs, intelligence agencies and conflict zones around the world, demanding urgent, coordinated action before the next generation of AI outpaces our ability to contain it.
