Anthropic disclosed Thursday that it had shut down several attempts this year by scientists to use its artificial intelligence models for research that could contribute to biological weapons development.
Anthropic published a report detailing misuses of its AI systems over the past eight months, describing cases in which state-linked actors and researchers tried to exploit its technology. Anthropic said it could not always tell whether the biological research in question was legitimate science or something more dangerous, but it banned the accounts involved regardless.
In some cases, according to the report, scientists got around Anthropic’s controls that block access from restricted regions and worked to “obfuscate the purpose of their research to evade our safeguards.” Anthropic did not name the researchers, their institutions or their countries, nor did it specify which biological agents were involved.
Jacob Klein, Anthropic’s head of threat intelligence, said the blocked cases rarely looked like an obvious threat. “You are not seeing someone in a comic book kind of way say, ‘Hey, I want to build a biological weapon to kill everybody.’ It’s an incredibly nuanced situation,” he said in an interview.
One case from May involved a scientist who asked Claude for help writing a grant application for research on chikungunya, a mosquito-borne disease that can cause months of severe joint pain and other symptoms. The scientist appeared to be seeking to engineer mutations that would make the virus more harmful through repeated infection of live animals, a method known as gain-of-function research. Anthropic said the work appeared to be linked to a military research institute.
“What we don’t know is if the research was meant to be weaponized. But a military institution doing gain-of-function research is concerning,” Klein said.
The report also described government-linked actors from China and Iran using Claude to help surveil dissident and diaspora communities. Separately, Anthropic said Russian state media had allegedly used the model to produce online propaganda dressed up as independent journalism, including fabricated claims tied to an election in Moldova.
Anthropic also identified a new category of abuse: attempts to use Claude to help develop conventional weapons, including firearms, missiles, armed drones and bombs. The report cited three such cases in China, two in Russia and one in Yemen. Anthropic did not name the group behind the Yemen case, though the report’s context points to the Iran-backed Houthi militia.
The Yemen group’s work touched several missile programs at once, according to the report. That included a multi-stage ballistic missile with a stated range of more than 2,000 kilometers, and a missile system called the “R2000” set, which had a hypersonic glide vehicle variant. Anthropic said the actors used Claude Code “in place of human software engineers” to write guidance, navigation and control software for flying vehicles, and used the model to integrate an open-source autopilot with a phone-class flight computer. The work covered control and position-estimation software, tuning control settings, running firmware builds and conducting flight simulations. To move faster, the group ran multiple Claude instances at once: one to write code, one to research, and a third to review the first instance’s output.
Anthropic found no evidence that an operational weapon was ever successfully fielded. The actors did carry out a test launch of a guided rocket, which appeared to fail, and within hours they were back in Claude asking it to help investigate what went wrong. The group had already built an offline simulation toolkit on its own, independent of Claude or standard engineering software such as MATLAB, before it ever turned to the AI model for help.
The Yemen case was one of four conventional weapons operations detailed in the report. The others involved software tied to a drone swarm, a torpedo-interception system, and targeting software for electronic warfare and suppression of air defenses.
The company said older AI models released last year weren’t capable enough to meaningfully assist with dangerous biological research. Newer, more capable models have pushed it to tighten safeguards around dual-use biological queries.
AI giants simultaneously claim their AIs are intelligent yet demonstrate vulnerability to foreign actors. Breitbart News social media director Wynton Hall has written his instant bestseller Code Red: The Left, the Right, China, and the Race to Control AI to serve as the definitive guide on how the MAGA movement can create positions on AI that benefit humanity without handing control of our nation to the leftists of Silicon Valley or allowing the Chinese to take over the world.
Lucas Nolan is a reporter for Breitbart News covering issues of AI, free speech, and online censorship.
Read the full article here