In brief
Anthropic has identified and disrupted a series of attacks that used its Claude family of models to aid biological weapon development, revealing a growing threat to bio‑security. The firm’s threat‑intelligence report documents five case studies involving malicious actors, as well as six instances where AI was used to design conventional weaponry. With the risk highlighted by safety researchers and political figures, Anthropic stresses that the same technology that can build deadly pathogens could also forge cures, underscoring the need for stronger safeguards. The company has shared its findings with law‑enforcement partners and is enhancing its internal controls to prevent future misuse.
Anthropic, the creator of the Claude family of models, has released its first threat‑intelligence report of 2026, revealing a wave of hostile actors seeking to use its AI for bio‑weaponry.
The firm says it identified and disrupted several attempts to obtain step‑by‑step instructions for producing weaponised pathogens. It also flagged the deployment of its models in state‑backed cyber‑espionage and propaganda campaigns.
According to spokesperson Jacob Klein, “the same information that can be used to make a biological weapon could also be used to develop a vaccine or cure for a disease.” He stressed that even a small fraction of misuse could have catastrophic outcomes.
The report lists five case studies where Claude aided biological weapon development, and six where it was used to design conventional‑weapon software, highlighting the dual‑use dilemma of frontier AI.
As the industry faces growing calls for pause and regulation following warnings from US senator Bernie Sanders and safety researchers, Anthropic says it has shared its findings with authorities and is tightening its safeguards.
Read the full report on Anthropic’s website: Full Threat‑Intelligence Report.
