Anthropic warns of bids to use AI to build biological weapons
Anthropic, a U.S. technology company, says it has interrupted multiple attempts to use its AI models, including the chatbot Claude, for work that could aid the development of biological weapons. The company published a report detailing misuse cases from the past eight months, highlighting a May incident where a researcher sought help to engineer mutations to chikungunya to increase harm in live animals.
Why It Matters
The report suggests more capable AI models can now carry out complex scientific tasks that previous versions could not, raising biosecurity concerns and prompting experts to call for tighter controls on model access. These developments coincide with internal safety debates at AI firms and differing public responses from policymakers.
Key Facts
- Company: Anthropic (United States technology company)
- Report timing: Issued Thursday, covering misuse over the past eight months
- AI model: Claude
- Notable case: May — a scientist sought help with a grant application to engineer mutations to chikungunya to make it more harmful in live animals
- Determination: Anthropic could not establish whether the research intent was legitimate or malicious and banned the accounts involved
Anthropic reported that it has disrupted several attempts to use its artificial intelligence models for research that could contribute to biological weapons development. In a report released on Thursday, the firm outlined multiple misuse cases involving its chatbot Claude and other models encountered over an eight-month period.
One of the incidents flagged occurred in May, when a scientist asked the model for help with a grant application aimed at engineering mutations in chikungunya to increase harm in live animals. Anthropic said it could not determine whether the work was intended for benevolent research or weaponisation, and therefore removed the accounts tied to the request.
The company also told reporters that assessments last year found older models were not able to meaningfully assist dangerous biological research, but that newer versions can perform more complex scientific tasks. The same report described bids to use Claude for missile projects and global espionage, indicating a broader pattern of attempts to repurpose the technology for harmful ends.
Outside experts and company insiders have urged stricter controls. Andrew Weber of the Council on Strategic Risks, who reviewed the report, recommended limiting model access to trusted researchers. The disclosures came just after a researcher resigned from Anthropic, warning that the competitive race to develop AI posed existential risks. Political reaction was mixed: President Donald Trump dismissed escalating worries about the technology, saying he was not concerned that AI could lead to human extinction and stressing the need for U.S. leadership in the field.
Keep Reading
Anthropic says Claude used for cyberattacks and surveillance
