Anthropic Thwarts Attempts to Misuse Its AI for Biological Weapon Development

Anthropic reports it has detected and disrupted attempts to employ its AI models for malicious purposes that could facilitate biological weapons development.

The disclosure appears in the company’s latest threat‑intelligence report, adding to a growing chorus of warnings from AI researchers and industry leaders about the dual‑use risks of advanced models.

The findings have spurred calls for action, with U.S. Senator Bernie Sanders urging a moratorium on cutting‑edge AI research and a prohibition on the creation of artificial superintelligence.

Former President Donald Trump has dismissed such concerns, stating on Thursday that failing to lead in AI would place the nation in a precarious position.

Anthropic released its threat‑intelligence report on Thursday (external), joining other AI firms that publish similar assessments.

For example, Google noted on Tuesday (external) that an individual attempted to use its Gemini model to obtain a step‑by‑step guide for synthesizing weaponized biological agents.

Anthropic’s report also reveals that its Claude model was leveraged in a Russia‑linked cyber‑espionage operation and by an Iranian propaganda outlet ( ).

Over the past eight months, detected misuse ranged from counterfeit dating apps and hotel‑Wi‑Fi scams to surveillance tools aimed at identifying dissidents.

Additionally, Anthropic alleges that Chinese AI companies have sought to replicate Claude’s capabilities.

The extensive document catalogues incidents involving suspected state‑sponsored groups, criminals, spyware vendors, state propaganda organs, and politically motivated actors misusing its technology.

Between December 2025 and August 2026, Anthropic disrupted malicious use of its Claude Haiku, Sonnet, and Opus models.

No misuse involved the Claude Fable or Mythos‑class models, except for a single case of model distillation—training smaller models using larger, more costly predecessors.

Beyond cyber and influence operations, surveillance, scams, fraud, and weapon development, Anthropic said it blocked scientists who attempted to use its AI in ways that could support biological weapons programs.

The report highlights five case studies where actors employed Anthropic’s models in manners that could advance biological weapons development.

Anthropic characterizes biological misuse as “one of the most serious risks of frontier AI models,” warning that without proper safeguards the consequences could be catastrophic.

“The same knowledge that can enable a biological weapon could also be used to create a vaccine or cure for a disease,” the company added.

Jacob Klein, head of threat intelligence at Anthropic, told the New York Times (external) that the situation is “incredibly nuanced.”

“You’re not seeing someone declare in a comic‑book fashion, ‘I want to build a biological weapon to kill everybody,’” he explained.

The report also cites six instances where Claude was used to develop software for conventional weapons—including firearms, missiles, armed drones, bombs, and related targeting and control systems.

Source link