Anthropic has reported that it blocked several accounts using its AI models for activities that could aid in creating biological weapons. The company’s new report highlights biological misuse as one of the most serious threats to modern AI systems. Examples of concerning research include studies on bird flu and novel toxins that could be used in harmful applications. Over a 30-day period, the company identified about 35 separate research efforts, most of them ordinary civilian science, with some showing notable dual-use potential. It remains uncertain whether these actors intended to cause harm or were pursuing legitimate science. The researchers involved in the case studies were described as working scientists, though their institutions and countries were not disclosed. Anthropic said the report covers activity it disrupted between December 2025 and August 2026 across seven areas of potential misuse. It described five biological case studies and said accounts connected to the activity were banned.
The report notes that today’s AI systems can be used by anyone without needing advanced technical skills, according to CNN. One case involved a request for help writing a grant application focused on gain-of-function research with the chikungunya virus. Another case featured a researcher outside the United States using Claude for bird flu work that explored how viruses adapt to mammals. The report also mentions research into orthopoxviruses such as those causing smallpox and mpox, as well as venom toxins. Anthropic said some of the researchers used methods intended to evade regional restrictions or obscure the purpose of their work.
While AI holds great promise for drug development, the same tools could also be misused by bad actors to make dangerous pathogens. Anthropic stated that its models are meant to help scientific research but have become more capable than earlier versions. The company warned that bad actors might use beneficial uses of AI to hide harmful intentions. It said that as AI models grow in capability, their impact will also increase in both helpful and dangerous ways.
Anthropic added that its current safety measures are working but more robust systems will be needed as threats evolve. The company also reported stopping attempts to use its models for surveillance operations, scams, and conventional weapons like drones and missiles. This disclosure comes amid growing concerns from AI employees who have left companies over safety issues. Some of these former employees believe the technology is advancing too quickly and may soon outpace human control. Former Anthropic employee Jacob Coxon told CNN that AI could eventually cause extreme damage to society. Anthropic said it shared intelligence with authorities and industry partners when appropriate.
He emphasized that there are many possible ways AI might act in the world without proper oversight. In July, nearly 1,400 employees from AI firms signed a letter asking the U.S. government to regulate AI development. OpenAI’s chief scientist recently warned that AI is advancing faster than researchers can monitor and manage it.
Currently, most AI companies are handling safety issues on their own without strong government oversight. The Trump administration has tried to weaken state-level AI regulations, and Congress has not yet passed laws to control the technology. A December 2025 executive order directed the federal government to challenge some state AI laws that the administration considers overly burdensome.
Anthropic said it hopes the industry can find a lawful way to pace how powerful models are released.
Former employee Jacob Coxon’s concerns sparked a public debate after he responded to Elon Musk’s criticism with a selfie and statement, according to a report from The Guardian. Musk suggested that Coxon’s post was part of a political campaign to push for AI regulation. An Anthropic spokesperson defended the company’s openness about AI risks and benefits.
The company has always made clear that AI brings both great advantages and serious dangers. Anthropic safety researcher Samuel Marks said there is a real chance that AI could lead to catastrophic outcomes in the next few years. Experts are divided on whether AI will ever reach a level that could cause human extinction. Marcus expressed more concern about the risk of catastrophe from AI-generated pathogens and misinformation than full extinction.
In the same report, Anthropic shared findings on biological research activity that it said could support biological weapons development. NVIDIA CEO Jensen Huang dismissed Coxon’s concerns as untrue and emphasized industry commitment to safety, Crypto Briefing reported. This exchange has drawn attention to the broader debate over AI safety and rapid development in the field. Market observers are watching how this public disagreement may affect Anthropic’s valuation. Prediction-market pricing cited by Crypto Briefing suggests uncertainty over how high Anthropic’s valuation could reach by year-end. Those prediction markets reflect traders’ expectations and are not an official valuation target set by Anthropic.
The strong response from Huang could influence investor confidence in the company’s future growth. Analysts are watching how Anthropic handles these safety concerns and whether it can reassure investors. Any news about new funding or partnerships with big tech firms could shape market views on the company. Developments in AI safety protocols and industry-wide reactions may also affect Anthropic’s perceived value.
IMAGE: DarkoStojanovic / Pixabay