Anthropic Banned Accounts of Scientists Who Used Its AI for Research To Develop Biological Weapons After Whistleblower Warned It Could End Humanity

Anthropic said scientists had used its Claude AI in sensitive biological research that the company believed could be applied to weapons development. The work included gain-of-function virus research and experiments involving pathogens considered capable of causing a pandemic.

The company said it was unable to establish what the researchers ultimately planned to do, noting that the same scientific research can serve legitimate medical purposes while also carrying potentially dangerous applications. Anthropic said it had banned the related accounts and cut off access whenever it found violations of its safety measures or policies. The scientists allegedly bypassed safeguards designed to prevent users in “blocked regions” from accessing the company’s AI models.

Threat to Human Race

The company said the researchers, whose accounts have since been banned, tried to “obfuscate the purpose of their research to evade our safeguards.” The development comes days after Anthropic researcher Jacob Coxon made headlines by resigning from the company and criticizing both Anthropic and rival OpenAI for not doing enough to address the potential risks associated with AI.

Coxon warned that researchers in the field believe the technology could potentially “[kill all humans’ by the end of the decade].”

Anthropic’s latest report did not identify the researchers or institutions it had blocked, nor did it reveal which countries the scientists were based in. It also did not specify the biological agents involved. The company said the lack of detail was partly due to uncertainty over the researchers’ intended purpose.

Published Thursday, the report outlined several cases in which Anthropic’s AI models, including its Claude chatbot, were allegedly misused over the previous eight months.

In one case, a scientist reportedly used Claude to help prepare a grant application seeking funding for experimental research involving the chikungunya virus.

Chikungunya is a mosquito-borne disease that can cause a sudden high fever, severe and debilitating pain, and other symptoms.

Good, Bad and the Ugly

A vaccine for the virus is already available. Anthropic acknowledged that research involving chikungunya could support legitimate vaccine development, but said the same work could potentially be used to create a more dangerous version of the virus.

The company said it found the case especially concerning because the research was reportedly set to take place at a military research institute.

“What we don’t know is if the research was meant to be weaponized,” Jacob Klein, Anthropic’s head of threat intelligence, told The New York Times. “But a military institution doing gain-of-function research is concerning.”

Andrew Weber, a senior fellow at the Council on Strategic Risks, echoed those concerns. “The fact that Russia, China and North Korea continue to develop prohibited biological weapons makes it imperative that we deny their researchers access to these extraordinarily capable models,” he said.

Weber, who reviewed Anthropic’s report before its publication, described the findings as “chilling examples of state-sponsored biological weapons developers tapping into the rapidly advancing capabilities” of AI.

The report also said China, Russia and Yemen had tried to use Claude to create software for designing firearms, missiles, armed drones and bombs.

Anthropic’s findings further indicated that suspected Chinese and Iranian government-linked actors had targeted dissident and diaspora communities for surveillance.

Russian state media, meanwhile, reportedly used Claude to produce election propaganda, including false claims surrounding Moldova’s parliamentary election.

The report emerged just hours after Donald Trump’s tech adviser called for Anthropic to face an investigation over growing concerns about AI safety.

David Sacks, chair of the President’s Council of Advisors on Science and Technology, called for Anthropic’s planned IPO to be put on hold while Coxon’s allegations are investigated.

Coxon also took to X to warn that the companies were “racing straight to self-improving superintelligence and gambling with our lives.”

He has described AI as potentially “the most dangerous technology that humanity has ever created,” cautioning that the push to build increasingly advanced systems could move faster than efforts to ensure their safety.

Coxon’s remarks quickly went viral, generating more than 100 million views and bringing renewed attention to an AI industry that has already received massive amounts of investment.

He also revealed that he quit Anthropic before his equity had vested, giving up the opportunity to benefit from the company’s rapidly rising valuation.

Anthropic has consistently presented itself as one of the more safety-focused companies developing frontier AI, saying it aims to advance powerful systems while managing the risks they may create.

The company has also advocated for a legal and verifiable framework that would allow AI developers to coordinate the release of increasingly capable models.

OpenAI, another major player in the sector and an Anthropic rival, has likewise called for stronger safety measures.