Anthropic has revealed that it blocked multiple attempts to use its Claude AI models in ways that could have supported the development of biological weapons — the most serious category of misuse identified in the company’s first threat intelligence report of the year, published Thursday.
The report covers malicious activity disrupted across Claude’s Haiku, Sonnet, and Opus model family between December 2025 and August 2026. It documents five case studies of actors using the models in ways that could support biological weapons development — a category Anthropic describes as “one of the most serious risks of frontier AI models.” None of the misuse cases involved Claude Fable or the more powerful Mythos-class models, with one exception involving distillation, the process of training smaller AI models using outputs from larger ones.
“Without the correct safeguards, such capabilities could have catastrophic consequences,” the company said. Biological misuse is particularly difficult to police because the information involved is dual-use. “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,” Anthropic noted. Jacob Klein, the company’s head of threat intelligence, described the situation to the New York Times as “an incredibly nuanced situation,” adding: “You are not seeing someone in a comic book kind of way say, ‘Hey, I want to build a biological weapon to kill everybody.'”
The Full Scope of What Anthropic Found
The report is wide-ranging and notable for its specificity. Beyond biological weapons, it documents six cases in which Claude was used to develop software for conventional weapons — including firearms, missiles, armed drones, bombs, and munitions, alongside the targeting and control systems that operate them.
Cybercriminals and state-backed hacking groups feature prominently. A group whose activity is consistent with the Russia-based Midnight Blizzard is identified as having allegedly used Claude to build a system that automatically detected when its malware was flagged by security defences and then rewrote the code until it evaded detection — a particularly concerning use case that points toward AI-assisted cyber capability development. Hacking group ShinyHunters and China-based labs are also named in the report.
Russian and Iranian state actors were identified as misusing Claude as well. The report describes actors linked to a Russia-based cyber espionage campaign and an Iranian state propaganda institution as having used the models. Anthropic said it had shared relevant intelligence with authorities and industry partners where appropriate.
The broader catalogue of disrupted cases spans fake dating apps, hotel WiFi scams, surveillance tools built to identify political dissidents, and coordinated influence operations. The company also accused Chinese AI firms of attempting to replicate Claude’s capabilities through distillation.
Anthropic said it had incorporated the report’s findings into its internal processes “to better prevent, detect, and disrupt these activities in the future.”
An Industry-Wide Pattern
Anthropic’s disclosure is part of a broader pattern across the AI industry. Google published a blog post on Tuesday describing an attempt to use its Gemini model to obtain “a complete, step-by-step technical guide for synthesizing weaponised biological agents.” Such threat intelligence reports have become a regular feature as AI companies seek to demonstrate how they identify and respond to misuse — though the frequency and severity of what is being reported has increased markedly over the past year.
The report arrives as the AI safety debate intensifies at both corporate and governmental levels. Anthropic’s own top safety researcher has stated publicly that he believes there is a greater than 10% chance AI could kill all humans within the next decade. OpenAI chief scientist Jakub Pachocki called last week for the industry to implement “voluntary slowdowns” until safeguards are in place. “I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence,” Pachocki wrote.
Those warnings have prompted political responses. In the UK, an open letter to Prime Minister Andy Burnham called for a new multinational treaty on safe AI development. In the United States, Democratic Senator Bernie Sanders has introduced legislation to ban AI superintelligence and temporarily pause advanced AI development. “When scientists tell you there is a chance, a chance that it could have a cataclysmic impact on humanity, you have to be a moron not to say, slow it down,” Sanders said Thursday.
President Trump has taken a different position, expressing concern on Thursday not about AI’s risks but about the possibility of losing the AI race. “If we don’t win AI, we’re going to be put in a very bad position,” he said.
This article discusses AI misuse for weapons development. These are rapidly evolving safety and policy issues. Anthropic’s full threat intelligence report is available at anthropic.com.
Stay informed. Subscribe to the JournalTodays Newsletter for the latest AI safety news, cybersecurity coverage, and technology updates delivered straight to your inbox.
