Anthropic Blocks Malicious Use of AI for Weapons Development
Anthropic has announced that it successfully identified and thwarted multiple attempts to utilize its artificial intelligence model for “malicious activity,” including efforts aimed at developing biological weapons. These findings were detailed in the company’s recent threat intelligence report.
This report contributes to a growing body of warnings from AI experts and industry insiders regarding the inherent risks posed by advanced artificial intelligence technologies. According to the report, misuse of the company’s models—specifically Claude Haiku, Sonnet, and Opus—was detected and disrupted between December 2025 and August 2026. The misuse was attributed to a range of actors, including suspected state-sponsored organizations, criminals, spyware vendors, state propaganda institutions, and politically motivated individuals.
The Scope of Detected Misuse
Anthropic stated that its intervention covered various areas of concern, ranging from cyber and influence operations to fraud, scams, and weaponry development. The company also reported blocking scientists who were using its AI in ways that could support biological weapons research.
The report highlighted five specific case studies detailing how actors attempted to leverage the models for biological weapons development. Anthropic stressed that biological misuse represents “one of the most serious risks of frontier AI model,” noting that without proper protective measures, such capabilities “could have catastrophic consequences.” The company pointed out the dual nature of the technology, stating that “The same information that can be used to develop a biological weapon could also be used to develop, for example, a vaccine or a cure for a disease.”
In addition to biological threats, the report documented six instances where Claude was used to create software for conventional armaments, including missiles, bombs, firearms, armed drones, and their associated targeting and control systems. Furthermore, the company noted that its AI tool was used by actors linked to a Russia-based cyber espionage campaign and by an Iranian propaganda institution, while also accusing Chinese AI firms of trying to replicate Claude’s capabilities.
Jacob Klein, the head of threat intelligence at Anthropic, provided context, stating that the situation is “an incredibly nuanced situation.” He added that misuse is rarely as explicit as someone stating, “Hey, I want to build a biological weapon to kill everybody.”
Industry Calls for Global AI Regulation
The revelations from Anthropic come amid escalating global alarm over AI safety. Google had previously warned on Tuesday that an individual attempted to use its AI tool, Gemini, to acquire a “complete, step-by-step technical guide for synthesizing weaponised biological agents.”
The concerns are prompting calls for significant governmental and industry intervention. US Senator Bernie Sanders has proposed legislation aimed at temporarily pausing advanced AI development and establishing a ban on artificial superintelligence. Speaking on BBC’s Newsnight, Sanders asserted that if scientific experts warn of a “chance… cataclysmic impact on humanity,” it would be unwise to ignore that warning.
Other experts have also voiced serious concerns. A top safety researcher at Anthropic previously warned that AI is advancing so rapidly that there is a greater than 10% probability it “could kill all humans” within the next decade. Similarly, OpenAI chief scientist Jakub Pachocki called for the industry to implement “voluntary slowdowns” until comprehensive safeguards are established. Pachocki stated, “I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence,” while noting that broader interventions are required beyond OpenAI’s internal efforts.
The threat report also indicated that cybercriminals and state-backed hackers are increasingly using AI technology. The report named groups such as ShinyHunters and China-based labs, and detailed an instance where a hacking group consistent with the Russia-based Midnight Blizzard allegedly used AI to build a system that automatically detected and rewrote malware code to evade security defenses.
Anthropic confirmed that it has shared its findings and intelligence with relevant authorities and industry partners.
