Anthropic blocks scientists attempts to use AI for biological weapons research
Anthropic disclosed on Thursday that it had blocked multiple attempts by scientists to use its AI models, including Claude, for research that could support the development of biological weapons.
The company detailed the cases in a report covering misuse detected and stopped between December 2025 and August 2026. It said some activity involved researchers linked to states or operating in regions where access to its services was restricted, including China, Russia and Yemen. One case involved a scientist seeking help to prepare a grant application for gain-of-function research on the chikungunya virus, with proposed work including engineering mutations to make the virus more harmful. Another case involved a researcher outside the U.S. using Claude to plan experiments with highly pathogenic avian influenza and its adaptation to mammals.
Anthropic said its safety systems prevented the researcher from using its more capable models, limiting access to a weaker model with restricted assistance. The company noted that many biological research techniques can serve both legitimate and harmful purposes, making assessments difficult. It added that it banned accounts connected to the activity, shared relevant information with authorities and industry partners, and used the incidents to improve its safeguards.
Jacob Klein, Anthropic’s head of threat intelligence, told The New York Times that the situations were rarely straightforward, as users often did not openly declare harmful intent. The company said it had no public evidence of private firms sharing proof of AI platform misuse for biological weapons development .
Follow us for live European news
- 3
- 2
- 2
- 2
- 1
- 1
- 1
5 further sources not geolocated
Articles
Anthropic details how Claude was misused for surveillance and weapons Anthropic has published its most detailed account of Claude misuse, laying out how people tried to use its models for harm. The company said it detected and shut down these operations. They ran across cyberattacks, state surveillance, influence campaigns, conventional weapons work and biological research. It set out the cases in a threat intelligence report […] This story continues at The Next Web
the next web · about 2 hours ago
Anthropic afirma que ha bloqueado varios intentos de crear armas biológicas con su IA La empresa de inteligencia artificial Anthropic afirmó este jueves que ha frustrado varios posibles complots en los que se utilizaron sus modelos de Inteligencia Artificial (IA) para realizar investigaciones que podrían haber contribuido al desarrollo de armas biológicas. Leer
expansion · about 3 hours ago
Anthropic asegura que ha detenido varios intentos de crear armas biológicas con su IA La empresa tecnológica ha asegurado haber desarrollado sistemas de detección para identificar y bloquear los usos que puedan suponer un riesgo en sus modelos de IAUn ingeniero de Anthropic dimite y denuncia que las empresas de IA ocultan su riesgo real: Están apostando nuestras vidas La empresa de inteligencia artificial Anthropic ha advertido este jueves de que ha frustrado varios posibles complots en los que se utilizaron sus modelos de Inteligencia Artificial (IA) para realizar investigaciones que podrían haber contribuido al desarrollo de armas biológicas. La compañía ha indicado en un informe que detectó y bloqueó estos usos indebidos durante este año, aunque no pudo determinar si los estudios tenían fines legítimos o maliciosos, por lo que optó por bloquearlos ante el riesgo de que sus modelos pudieran contribuir al desarrollo de patógenos peligrosos. Si bien las evaluaciones son útiles porque proporcionan evidencia de capacidad, no pueden demostrar concretamente que dicha capacidad se utilizaría alguna vez para desarrollar armas biológicas en el mundo real, ha explicado la empresa en el documento. Hasta el momento, la corporación californiana ha negado que tuviera constancia de que alguna empresa privada haya compartido públicamente evidencia del posible mal uso de sus plataformas para el desarrollo de armas biológicas. Anthropic ha incluido además en otro comunicado una extensa lista de casos detectados en los últimos ocho meses en los que se habían usado sus modelos de IA de manera indebida, además del chatbot Claude, en China, Rusia y Yemen. Este informe surge entre crecientes preocupaciones de expertos en el sector tecnológico por el uso dual de la IA, ya que los modelos capaces de ayudar en investigaciones biológicas también pueden servir para actividades peligrosas, así como también de voces que alertan de los peligros derivados de la creciente autonomía de estos modelos. La empresa tecnológica ha asegurado haber desarrollado sistemas de detección para identificar y bloquear los usos que puedan suponer un riesgo en sus modelos de IA.
eldiario.es · about 3 hours ago
Opinie: AI-detectie als Pangram inzetten creëert een cultuur van wantrouwen
trouw · about 3 hours ago
Hoe schadelijk is het dat een premier AI gebruikt voor berichten op sociale media? Jetten reageert luchtig: Ja, net als driekwart van de Nederlanders gebruik ik af en toe AI Ineens duikt op sociale media weer zijn bijnaam Robot Jetten op, nu blijkt dat een aantal berichten van premier Rob Jetten is gegenereerd door AI. Volgens experts dreigt afbreuk aan het imago en de authenticiteit van een politicus die zoiets doet.
nrc · about 3 hours ago
Jeff Bezos Washington Post assures everyone talk of AI killing humans is nothing to fret about The news outlet, whose owner recently co-founded an AI company, instead warned about looming government overreach against Big Tech
independent · about 3 hours ago
Sam Altman pitched AI to run utility company security amid warnings about robot takeover Cybersecurity experts are sounding the alarm about vulnerabilities in critical infrastructure from weaponized AI
independent · about 3 hours ago
Anthropic says it stopped scientists potentially developing bioweapons with AI Start-up discloses five examples of when users circumvented controls and made efforts to obfuscate the purpose of their research
financial times · about 3 hours ago
Is AI Actually Going to Kill Us All? This week on Uncanny Valley, we dig into a former Anthropic researchers AI doomsday warning, the latest upgrades from Apples event, and the census report that claimed Trump won the 2020 election.
wired.com · about 3 hours ago
Dez perguntas sobre o alerta que pôs um prazo na humanidade Jacob Coxon, de 27 anos, demitiu-se da Anthropic, onde fazia investigação pré-treino dos modelos
observador · about 3 hours ago
Hoppar av från AI-jätten – Kan döda oss alla AI-jätten Anthropic har drabbats av två uppmärksammade uppsägningar.Och nu får avhopparna rätt – från oväntat håll.Vi tror verkligen att AI skulle kunna döda alla människor, skriver en av Anthropics AI-säkerhetschefer, på X.
expressen · about 3 hours ago
After a researcher at Anthropic resigned in prote, Joshua Lowe describes a harrowing scenario which tech titans believe is plausible
bluesky_The Times · about 3 hours ago
BREAKING: Anthropic says it halted potential plots by scientists who used its A.I. models for research that could aid biological weapons development and help engineer dangerous pathogens READ:
telegram_Insider Paper · about 3 hours ago
Bad actors in China and Russia are already weaponizing Anthropics AI Anthropic says foreign scientists, hackers and security services circumvented restrictions to use Claude for dangerous work, including possible bioweapons research.
politico.eu · about 3 hours ago
Hundreds of AI agents helped PaperCut attacker hit 395+ orgs, and some went off script Human operator: dont touch CIS orgs. AI agents: look a squirrel!
the register · about 3 hours ago
AI researchers back an extinction warning as the Pentagon rejects it A warning from a departing Anthropic researcher has drawn a wave of responses this week. Senior scientists at rival labs backed the claim that AI could threaten humanity, while a top Pentagon official rejected it. The reaction followed Jacob Coxons resignation from Anthropic. In a post on X, he said AI firms were racing straight […] This story continues at The Next Web
the next web · about 3 hours ago
Surfshark CTO: AI is making cyberattacks more convincing and harder to detect - The Baltic Times As artificial intelligence becomes an everyday technology, it is also giving cybercriminals increasingly powerful tools. Donatas Budvytis, Chief......
baltic times · about 3 hours ago
NRD Cyber Security: AI is accelerating the cyber race for attackers and defenders - The Baltic Times Russias war against Ukraine and the rapid adoption of artificial intelligence are reshaping the cybersecurity environment, forcing govern......
baltic times · about 3 hours ago
Anthropic asegura haber bloqueado una operación para desarrollar armas biológicas con su modelo de IA La corporación californiana negó que tuviera constancia de que alguna empresa privada haya compartido públicamente "evidencia del posible mal uso de sus plataformas para ese desarrollo" Leer
elmundo · about 3 hours ago
Anthropic says it blocked attempts to use AI for potential biological weapons Start-up discloses five examples of when actors circumvented controls and made efforts to obfuscate the purpose of their research
financial times · about 3 hours ago
Künstliche Intelligenz: Wird KI „alle Menschen töten? Das sagt das Silicon Valley zur Doomsday-Debatte Alarmisten gegen Skeptiker: Die Debatte über eine unverantwortliche KI-Entwicklung nimmt Fahrt auf. Forscher Nate Soares fordert im Handelsblatt ein sofortiges Eingreifen der Politik.
handelsblatt · about 3 hours ago