Foreign Governments Used Claude to Research Bioweapons

Take Action

People are using the AI chatbot Claude in ways that could help them design deadlier germs. Many of them are connected to foreign governments or militaries.

They've used tools to hide their location. This lets them use Claude from countries where it's not supposed to be available, such as Russia, Iran, or North Korea.

What They Researched

  • In May 2026, someone asked Claude to prepare research materials on chikungunya, a mosquito-borne virus. They wanted to make the virus spread more easily. The research was to be conducted at a foreign military institute.
  • Someone used Claude to help plan experiments with bird flu. They were studying how the virus spreads from birds to mammals and were especially interested in how it could cause severe illness. They used tricks to hide their identity and exchanged thousands of messages with Claude over several weeks.
  • Someone asked Claude to draft a grant application for lab research on viruses like smallpox. They wanted to find out which viral genes can shut down the immune system of an infected host. Claude did not block the request, worked on it for about an hour, and produced detailed research plans.
  • Someone gave Claude data on animal venoms and asked it to generate new ideas for similar chemicals. Some of the results it generated could cause paralysis or be used as bioweapons.
  • Someone used Claude to redesign toxins across several research projects. This included parts of pathogens that could be used to create a pandemic.

Getting Around Safeguards

In these examples, bad actors hid the purpose of their research to try getting around Claude's safeguards. They were caught, and their accounts were banned from Claude. Some of them simply registered new accounts or switched to using other AI providers.

This raises the question of what Claude's safeguards haven't caught. These risks are hard to control because biological research is "dual use." This means that knowledge to make a deadly disease is the same knowledge needed to develop its cure. Bad actors can exploit this by phrasing harmful requests to seem innocent to the AI.

Why It Matters

Experts have been warning that advanced AI could help bad actors create bioweapons or start a deadly pandemic. As AI is trained to be more powerful, this becomes more likely. Politicians are paying attention when people contact them to share these concerns.

Sources

Quotes

  • Anthropic: "Biological misuse is one of the most serious risks of frontier AI models. It has long been a concern that AI models might one day reach the level of capability where they can help to make existing pathogens more dangerous—or create entirely new ones."
  • Anthropic: "actors circumvented controls we impose to prevent users from unsupported regions accessing our models, and engaged in other efforts to obfuscate the purpose of their research to evade our safeguards."
  • Anthropic: "Biological capabilities are dual use: they can be used for beneficial or harmful purposes, and it is often difficult to distinguish between them. 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."