Anthropic says it blocked possible efforts to build biological weapons

techSep 10, 202618552

Anthropic published a threat report saying it disrupted several attempts this year to use its Claude models for biological‑weapons research and related weapon design. The company said it shut down work it could not determine was legitimate or nefarious after observing suspicious activity, and it described attempts that included biological research as well as conventional‑weapons software design. Anthropic detailed incidents involving criminal hacking gangs, Chinese security bureaus, Russia‑linked spies and Yemen‑based arms manufacturers that used fraudulent accounts to access advanced assistants, and it flagged government‑linked accounts among actors trying to exploit the system. The report also described a hacktivist campaign and a single French‑speaking actor who built a custom Rust scanner to find exposed API keys. News outlets including The New York Times, NBC News and Forbes covered Anthropic’s disclosures and cited examples in the report, such as cases identified in China, Russia and Yemen. Anthropic said the pattern of abuse prompted the company to expand monitoring and to block or terminate suspicious projects rather than allow uncertain research to continue.

Frank Figliuzzi

@frankfigliuzzi.bsky.social

Anthropic saw people using Claude AI to possibly build biological weapons.Trump and his minions in Congress are willfully failing to regulate: www.nytimes.com/2026/09/10/u...

1184h ago

savefrogs2.bsky.social2

Thanks for sharing this. Shocking and disgusting that Anthropic workers et al. place their own personal wealth and professional curiosity above the health and safety of everyone else in the world. Meanwhile, America has a cabinet that refers to "A-One" [steak sauce].

Maya-181

Thank you very much for this!

balls-balls.bsky.social1

We we reached the point where corporations are less corrupt than the republican government

Quoting this

graylmitchell.bsky.social7

"The potential for cutting-edge A.I. models to facilitate the development of known or entirely new biological pathogens is among the gravest concerns experts have about a technology that is developing so rapidly that even its leading architects doubt whether humans will be able to fully control it."

Bill Kristol

@billkristolbulwark.bsky.social

“Anthropic said it had disrupted several potential plots this year by scientists who used its leading artificial intelligence models to conduct research that could have helped develop biological weapons.” AI. What could go wrong? www.nytimes.com/2026/09/10/u...

1333h ago

Jules⚖️📨

@northjules.bsky.social

Anthropic says it has disrupted several cases in which scientists used its Claude AI models for sophisticated biological research that could potentially contribute to biological-weapons development. #TheCityBrief

563h ago

Jules⚖️📨8

This is one of the clearest real-world examples yet of a risk AI companies have been warning about for years. And this time, Anthropic says its safeguards encountered actual users—not merely hypothetical scenarios in a laboratory evaluation.

TheBaroness 🐸7

Oh, great. Another day in fucking paradise.

1950studebaker.bsky.social3

And we’re relying on the billionaires, possibly the most evil group of people on our planet, to take care of all of this for us! Soon the greedy assholes will be asking for even more billions and trillions of taxpayer dollars to stop what they have created.

Alejandra Caraballo

@esqueer.net

I'm sure people trying to use Anthropic's models to potentially create bio weapons is all marketing. /s www.nytimes.com/2026/09/10/u...

943h ago

Alejandra Caraballo50

It's all just hype and marketing when rogue scientists can use advanced AI systems to potentially create bio weapons. The potential release of horrifying new viruses and diseases is totally just an attempt to get you to buy into AI and hype it up.

Sagittarius A*7

if it isn't marketing why is it in the NYT

Renezuo(OPEN Comms)5

With respect, yes, it sorta is. The marketing suggests that AI will somehow do this on its own unless you give more money to the company that made it so they can be the first past the post and avoid this tragic fate they created. A solution without a profit motive would be far simpler.

W./E. Ⓥ🐣🌻🌐✌️🗽⚖️🇵🇸5

I trust the NYT as far as I can throw their building, but I hear you

Dr. Alon Altman 🏳️‍⚧️3

Not all that anthropic does is marketing but this announcement definitely smells of marketing. It's very easy to get this kind of headline and it is mostly meaningless. I'm sure many security researchers tried to see if the models would do tasks related to bioweapons. That's all there is.

Professor Furious3

I think the key here is "Anthropic says".

Quoting this

ralphbunchonumbers.bsky.social2

I'm sure the press release from Anthropic about how powerful its product is has nothing to do with marketing ... /s

Cheryl Rofer

@cherylrofer.bsky.social

1. This is bullshit. 2. It is cause to shut down all the AI companies NOW. 3. Those companies might think about their pitch that "AI will take all your jobs and then kill you." 4. Those companies need new leadership. 5. Congress needs to act NOW.

564h ago

C.V. Danes2

More likely the AI will take all your jobs and then the cultist billionaires will kill you :-)

thomasselleck02.bsky.social1

I'm not sure how Fauci takes hits for supposedly being involved in " gain of function" research, and yet these AI folks get a pass on all this crap?

Hypervisible

@hypervisible.blacksky.app

“Anthropic is seeing AI move from an experimental tool into the everyday bureaucracy of state surveillance.”

323h ago

Quoting this

samuel2

“In Mali, a single consultant working for national security authorities used Claude to create a system that collects data from the country's mobile operators and builds dossiers about people.”

Ben Goggin

@bengoggin.bsky.social

Government-linked accounts tried to use Claude for work that could lead to bioweapons, Anthropic says www.nbcnews.com/tech/tech-ne...

44h ago

Quoting this

Paul Graf3

This, not AI agents propagating throughout the internet doing evil things, is where I think the greater danger lies. It always gets back to humans wanting to do unspeakable things to other human beings.

Marty Lederman

@martylederman.bsky.social

Better than the alternative, I suppose ... (now what are the odds that all of the companies, here and abroad, also realize 100% detection?) www.nytimes.com/2026/09/10/u...

114h ago

Representative Becca Balint

@balint.house.gov

Congress must urgently hold hearings on frontier research. The leaders of these massive AI companies owe us transparency about what they're doing. We can’t afford to continue the Trump strategy of letting billionaires play with the safety of humanity.

92h ago

Michael Mechanic

@michaelmechanic.bsky.social

Anthropic's guardrails won't suffice. We need a global pact that severely restricts access to any model with such capabilities. We're at not-unsubstantial risk, even without AI, of pandemics that would make covid look like child's play.

84h ago

Deborah Pearlstein

@debpearlstein.bsky.social

If you didn't see this coming, you don't have eyes. Credit of sorts to Anthropic at least for publicly reporting key examples (e.g. a foreign military research institute using Claude to help make a horrible mosquito-borne virus more harmful).

54h ago

Rob Bluebird

@bobbradysd1.bsky.social

Anthropic Says It Blocked Possible Efforts to Build Biological Weapons In a new report, the A.I. start-up added that it could not determine if the research was legitimate or nefarious, leading the company to shut down the work .. www.forbes.com/sites/antoni...

43h ago

Emily Feng

@emilyzfeng.bsky.social

“Anthropic documented another genre of abuse it said was new: attempts to use Claude to develop software for conventional weapons design and development… It detailed three cases in China, two in Russia and one in Yemen.” www.nytimes.com/2026/09/10/u...

65h ago

Jeffrey Vagle

@jvagle.me

This game is getting old, AI* tech bros. At what point does the let's-freak-out-regular-folks tactic stop working as an investment booster? www.nytimes.com/2026/09/10/u...

34h ago

Mikayla 🏳️‍⚧️🏴🦀

@rad.gendervibes.online

> we provide details of a hacktivist campaign […] > In the spring of 2026, a single French-speaking actor was observed using Claude […] > This actor built their own custom Rust-based scanner designed to scan and validate public containers for exposed API keys wow www.anthropic.com/threat-intel...

42h ago

1 source

nytimes.comAnthropic Says It Blocked Possible Efforts to Build Biological Weapons

Anthropic says it blocked possible efforts to build biological weaponstech Apple unveils foldable iPhone Duotech Apple unveils iPhone 18 Pro and 18 Pro Max with A20 Pro chiptech

Anthropic says it blocked possible efforts to build biological weapons

Related