Anthropic Blocks AI Attempt to Create Biological Weapons
Left 50%
Center 50%
Right 0%
1 left · 1 center · 0 right
What happened
Anthropic, the company behind the Claude AI models, said Thursday in a 154-page report titled "Detecting and Countering Misuse of AI" that it had identified five cases in which users sought the models' help with biological research that could support weapons development. The company said it could not determine whether that research was intended for legitimate medical purposes or for harm, and it banned the associated accounts, shared findings with government authorities and other AI labs, and strengthened its safeguards. One case involved a scientist asking Claude to help draft a grant proposal for gain-of-function research on the mosquito-borne chikungunya virus—work meant to identify mutations affecting transmissibility and immune evasion and engineer those changes into the virus—at a military research institute; other cases involved bird flu, orthopoxviruses, and non-transmissible venoms and toxins. Anthropic said some of those users routed prompts through third-party platforms to evade regional blocks after requests were rejected, and that it worked with partners to take down those relay networks. The report covered activity the company said it disrupted between December 2025 and August 2026 and separately described Iran-linked influence, surveillance, and U.S. naval-targeting activity as well as other criminal and state-linked misuse.
Sourcing check · Mostly on the record · 6 articles
4 of 5 articles cite a named source for the story. It started with unnamed sources; a named source appeared 4 hours later.
What these terms mean and how they're counted
- On the record · names a source 4 of 5
- On the record means the article cites a source by name for the story's main claim: a person, or an institution speaking officially, such as a company statement or a named agency's spokesperson. The opposite is an unnamed source. The count is how many of the counted articles cite at least one such source; the names are everyone named across them, each once however the articles spell it. Here: Anthropic and Evan Hubinger
- Primary evidence · cites a document or data 4 of 5
- Primary evidence means the article cites a record you could check yourself: a court filing or ruling, a dataset or report, a recording or transcript, an official statement or press release, market data. A quote is not a document. The count is how many articles cite one. Here: company report, threat report
- Credits another outlet 0 of 5
- The article says the story comes from another outlet's reporting (“first reported by Reuters”). Those articles relay the story; they are not independent confirmation. Here: Each article presents its own reporting.
- Unnamed sources 0 of 5
- The main claim is attributed to people who are not identified: “people familiar with the matter”, “officials said”. The count is how many articles do this.
- How it developed
- When the first report landed and whether it named anyone; when the first named source and the first document appeared. Here: First report Sep 10, 17:36 UTC by New York Post, with no named source. First named source 4 hours later, from New York Times. First document 4 hours later, from New York Times (company report).
- Not counted 1
- Opinion, analysis and video pieces, and articles too short to assess. Every count above is out of the counted articles. An article republished under a new headline is one article.
This is about who is cited, not whether the story is true. Each source below says what it names. How we count
Omitted — what each side leaves out
Unpacked
Left-leaning coverage never names the experiments Anthropic actually blocked. Right-leaning coverage reports a grant for gain-of-function research on chikungunya that Anthropic flagged in part because the work was meant for a military research institute, a detail carried by the New York Post. BBC says only that unspecified cases could support biological weapons and that the company could not tell legitimate science from harm. Without those details the story reads as a thwarted weapons plot rather than dual-use lab work Anthropic itself refused to classify.
Left-leaning coverage calls the blocked use "malicious activity." Right-leaning coverage describes the same users as scientists doing dual-use research whose intent could not be established.
Which country ran the military research institute that sought the chikungunya grant?
Bottom line
If you read only BBC, Anthropic stopped possible biological-weapons work. The New York Post is where that same shutdown is a chikungunya gain-of-function grant written for a military research institute.
The receipts — 6 facts one side's coverage lacks
- One blocked case was a grant application for gain-of-function research on chikungunya intended to take place at a military research institute.
- Researchers used a third-party platform to bypass regional restrictions and automatically route rejected prompts to another AI model.
- Other blocked biological cases involved bird flu, orthopoxvirus research, and non-transmissible venoms and toxins.
- An Iran-nexus actor used Claude to develop targeting recommendations against US naval forces.
- Anthropic disrupted six cases in which Claude was used to develop software for conventional weapons, including firearms, missiles, armed drones, and bombs.
- Google reported a person had attempted to use Gemini to obtain a complete step-by-step technical guide for synthesizing weaponised biological agents.
The Left View
Left-leaning coverage framed the report as Anthropic blocking "possible" efforts to use AI for biological weapons and as the latest in "a growing series of warnings" about the technology's "potential risks to humanity." The BBC highlighted Anthropic's statements that biological misuse is "one of the most serious risks of frontier AI model" and that without safeguards such capabilities "could have catastrophic consequences," and quoted threat-intelligence head Jacob Klein calling it "an incredibly nuanced situation" in which no one is saying, in "a comic book kind of way," that they want "to build a biological weapon to kill everybody." That coverage placed the findings against Senator Bernie Sanders's demand for a pause on advanced AI development and a ban on superintelligence, and against President Donald Trump's rejection of those fears, including his remark that "if we don't win AI, we're going to be put in a very bad position." The BBC also situated the report after a top Anthropic safety researcher's warning of a greater than 10% chance AI "could kill all humans" within the next decade.
The Right View
Right-leaning outlets stressed that the blocked users were scientists whose dual-use work included the military-institute grant case, and that Anthropic "could not determine the researchers' ultimate intent" and was not alleging they intended to cause harm. Fox Business headlined the biological cases together with "Iran-linked cases" and quoted the company saying sophisticated actors use biology's dual-use nature to maintain "plausible deniability," including through the relay networks used to evade blocks. The New York Post presented the findings as "real-world context" for former Anthropic researcher Jacob Coxon's warning of an extinction-level threat and his claim of a "substantial probability that this technology could kill everyone."
Our Take (balanced)
Left-leaning sources' strongest argument is Anthropic's conceded inability to distinguish vaccine-relevant research from weapons-relevant research, joined to its own warning that biological misuse "could have catastrophic consequences," which they read as proof that the danger is already live and that figures such as Sanders are reacting to that gray zone. Right-leaning sources' strongest argument is the concreteness of what was blocked—the military-institute proposal, the bypass infrastructure, and Iran-nexus targeting—together with account bans and the company's refusal to allege harmful intent, which they treat as evidence of real dual-use and adversarial pressure rather than abstract catastrophe. The central unresolved tension is whether those disruptions show that providers can police dual-use biology, or that the same opacity that left Anthropic unable to determine intent means the true purpose—and the size of the risk—cannot be known from the activity itself.
4 sources
- Anthropic Says It Blocked Possible Efforts to Build Biological Weapons
- Anthropic blocks possible attempt to use AI to make biological weapons
- Anthropic says scientists used its AI for research that could aid biological weapons development
- Anthropic says it blocked possible efforts to use AI for biological weapons development, Iran-linked cases
More stories
The week's bottom lines, in your inbox
One email a week: the five stories that mattered and what they actually mean. Free.