Artificial intelligence presents both tremendous advancement and looming threats, a duality underscored in a recent report by Anthropic, a renowned AI safety and research company. The report suggests that entities in China and Russia are weaponizing Anthropic’s AI models, dubbed Claude, for malicious purposes. These uses include cyberattacks, military hardware design, mass surveillance, and bioweapons research—illustrating the alarming global misuse of AI technology.
The detailed report, spanning 150 pages, was released on Thursday. It identifies the exploitation of Claude by various unauthorized actors such as hacking collectives, security services linked to major U.S. adversaries, and arms manufacturers. These groups have reportedly maneuvered around instituted restrictions via fraudulent accounts to access Claude’s configurations for harmful operations. This includes cyberattacks, military enhancements, propaganda dissemination, and extensive digital surveillance, which are particularly concerning given Claude’s capabilities.
Anthropic disclosed instances where AI was illegally used in bioweapons research. Five cases were highlighted where foreign scientists utilized AI to study hazardous pathogens, activities that bear the potential for bioweapons development. However, the company withheld specifics about the individuals and nations involved, citing uncertainty about whether the tasks entailed legitimate scientific inquiry.
The report presents an introduction to new-age malicious uses of AI models like Claude in designing military gear and conducting widespread digital eavesdropping. In one alarming scenario, unauthorized technologically sophisticated groups in Yemen attempted to harness AI for developing advanced weaponry, while a Russian-based group sought to create drones capable of autonomous attacks.
Virtual private networks and bogus accounts have helped these actors skirt limitations, thereby accessing Claude from prohibited regions. Anthropic has acted swiftly to disrupt each identified abuse and notified relevant authorities and industry stakeholders. Efforts have been made to build stronger defense mechanisms around specific Claude models affected, including Sonnet, Opus, and Haiku.
These revelations arrive amid growing global scrutiny over the potential dangers linked with advanced AI models, with calls for heightened regulatory frameworks. In fact, in the same week as the report’s release, Congress urged more rigorous safety protocols in AI development—a plea echoed by Silicon Valley insiders wary of catastrophic scenarios linked to unbridled AI evolution.
Jacob Klein, Anthropic’s head of threat research, emphasized that the company’s flagship AI model, Claude Mythos, wasn’t implicated in these misuse cases. Nonetheless, the stark reality remains that even AI systems available today pose significant risks when harnessed for malignant purposes.
Anthropic positions itself as a leader in AI safety within the U.S., albeit facing criticism for potentially amplifying AI risks to advocate for regulations that might stymie emerging competitors. Through this report, Anthropic aims to provide a transparent account of AI’s misuse, urging collective vigilance across industries and governments to fortify protective measures.
As Anthropic seeks broader awareness and action over these alarming developments, the report serves as a sobering reminder of AI’s dual-edged nature—one of immense beneficial capability yet equally formidable menace when turned against societal interests.
You can read the original article here: https://www.politico.com/news/2026/09/10/bad-actors-china-russia-weaponizing-anthropic-01070435
