By Eduardo Baptista and AJ Vicens
Sept 11 (Reuters) – Anthropic said in a threat intelligence report on Thursday that several actors had used its Claude AI models for activities ranging from weapons development and cyber operations to surveillance and fraud.
Here are some of the most notable cases described in the report as alleged by Anthropic.
CONVENTIONAL WEAPONS
• CHINA-TAIWAN AIR DEFENSES: Anthropic said a China-based actor used Claude to develop an electronic-warfare and air-defense suppression software suite that ranked targets and modeled radar jamming. During the project, the actor changed a simulation to include 12 targets in Taiwan, including early-warning radar, Patriot and Tien Kung missile batteries, air bases and a command bunker. Anthropic said account information linked the actor to Chinese research institutions including the People’s Liberation Army Academy of Military Sciences. Anthropic banned accounts linked to the actor.
China’s foreign ministry said it was not aware of the Anthropic report and that the government maintains that AI should be developed for good and opposes distortion of facts and smears against the country.
China claims democratically governed Taiwan as its own territory and has never renounced the use of force to bring the island under its control. Taiwan rejects Beijing’s sovereignty claims.
• CHINA ANTI-TORPEDO SYSTEM: Anthropic said another China-based actor used Claude to help develop specifications and fire-control software for an anti-torpedo system intended for the Chinese navy. The actor used Claude to produce a technical proposal of more than 200 pages, compare the system with U.S. Navy technology and simulate hostile technical reviews to improve the proposal. “We assess the actor was associated with a Chinese defense industry manufacturer aiming to produce a weapons specification and acquisition proposal for the People’s Liberation Army Navy,” Anthropic said.
• CHINESE RESEARCH INTO MICROWAVE WEAPONS: A China-based defense-intelligence actor used Claude to research foreign high-power microwave weapons, identify components and suppliers and trace supply chains, Anthropic said. The actor sought information that could help reverse-engineer the weapons and develop countermeasures, and drafted restricted briefings for senior Chinese Communist Party, military or state-security officials, the company said.
• YEMEN MISSILE DEVELOPMENT: Anthropic said a cell of threat actors in northern Yemen used Claude to help develop software for a guided rocket, a planned ballistic missile with a range of more than 2,000 km and a missile variant incorporating a hypersonic glide vehicle. The actors used Claude for coding, simulation and troubleshooting, including after a guided-rocket test appeared to fail. Anthropic said it had no evidence they successfully fielded an operational weapon. “Our safeguards blocked many of their requests, but not all of them,” Anthropic said.
The disclosure comes as the Iran-aligned Houthis, who control most of northern Yemen, have intensified attacks on Saudi Arabia and its energy infrastructure. The Anthropic report did not identify whether or not the threat actors were Houthis.
• RUSSIAN DRONE SWARM: Anthropic said likely freelance Russia-based actors used Claude to develop software for an autonomous swarm of first-person-view attack drones, including terminal guidance, target selection and coordination between multiple aircraft.
Small FPV attack drones have become a central weapon in the war in Ukraine, where Russia and Ukraine have deployed them on a vast scale for reconnaissance and strikes.
The Russian Embassy in Washington did not respond to questions about specific allegations made by Anthropic or the report overall.
• RUSSIA DUAL-USE COMPONENT PROCUREMENT: Anthropic said a Russia-based procurement manager used Claude to identify intermediaries in China and Hong Kong to acquire European-made goods with potential military uses and work out routes through third countries that could obscure their destination.Items sought included German-made magnetometers, space-grade photovoltaic wafers and aviation oxygen systems, Anthropic said. Russia has increasingly relied on intermediaries and alternative trade routes to obtain foreign technology since Western countries imposed sweeping sanctions and export restrictions following its invasion of Ukraine.
BIOLOGICAL RESEARCH
Anthropic detailed five case studies of people attempting to use “our models in ways that could support biological weapons development.”
The attempts included efforts to get Claude assistance on drafting a grant application for funding research on modifying the chikungunya virus in a location where Anthropic does not offer its services. In another case, also in a location where the company does not operate — requiring users to obfuscate their location to utilize the services — Anthropic said it detected efforts to research highly pathogenic avian influenza. Other examples included research related to orthopoxvirus and non-transmissible novel venoms and toxins, Anthropic said.
CYBER OPERATIONS
• CHINA CYBER OPERATIONS: Anthropic said Chinese-speaking operators likely based in Changsha, the capital of central China’s Hunan province, used Claude in cyber operations against about 50 organizations, including foreign government networks.The group, which Anthropic said included two university students in Hunan, used AI-driven workflows to search for previously unknown software vulnerabilities, develop exploits and conduct parts of intrusions with limited human supervision.
• RUSSIA INTELLIGENCE COLLECTION SUPPORT: One hacking group allegedly ran phishing, hotel Wi-Fi hijacking and WhatsApp-takeover operations against targets in the Ukrainian government, military and diplomatic sectors, using AI at nearly every stage, Anthropic said.The group’s tradecraft was consistent with Russia-based threat actor Midnight Blizzard, which the U.S. government has previously linked to Russia’s SVR foreign intelligence service.
• IRANIAN NAVAL RECONNAISSANCE: Anthropic said it identified and disrupted an Iranian-linked operator that used Claude to collect and analyze publicly accessible data to develop targeting recommendations against U.S. naval forces in the region. The compiled material included a roster of U.S. personnel scraped from captions on public military photographs, publicly accessible ship and aircraft transponder identifiers, and commercial satellite-imagery query scripts.
The operator also directed Claude to compile software vulnerability research on shipboard systems related to ship satellite and other communication equipment and industrial control products.
SURVEILLANCE
• UYGHUR RECRUITMENT IN SYRIA: Anthropic said a China-aligned actor used Claude to track and approach Uyghurs in Syria, including people assessed as financially or personally vulnerable, and offer payment for information about Uyghur units that had joined Syria’s newly formed army.Beijing regards Uyghur armed groups abroad as terrorist organizations and has long cited militancy as justification for security policies targeting the predominantly Muslim ethnic group. Anthropic assessed with low confidence that the operator was a contractor working for Chinese state security.
• CHINESE RELIGIOUS AND POLITICAL SURVEILLANCE: Anthropic said China-linked actors used Claude to compile intelligence on targets including Catholic cardinals, Taiwanese Christian leaders, Tibetan Buddhists, dissidents and activists.
In a separate operation, Claude was used to automate government-style reports monitoring Uyghurs, Tibetans, Taiwan political figures, labor and student activists and foreign media, Anthropic said.
China tightly controls organized religion and political dissent and has been accused by Western governments and rights groups of conducting surveillance and pressure campaigns against critics overseas. Beijing has rejected accusations of transnational repression.
• COMMERCIAL SURVEILLANCE AS A SERVICE: Anthropic said it banned an account that used Claude to build a commercial surveillance platform to analyze, classify and profile social media activity of users in Iran and the Gulf region. Anthropic said its findings corroborate a February 2023 investigation by journalism network Forbidden Stories that documented a service pitched by a company called S2T Unlocking Cyberspace.
Anthropic said it identified the activity in its pilot stage and found no evidence that later stages of the company’s surveillance chain were used against real targets before the account was banned. S2T did not immediately respond to a request for comment.
• IRANIAN PROFILING OF ISRAELI AND JEWISH DIASPORA ORGANIZATIONS: Anthropic said it identified an Iranian-linked operator that used Claude to build an automated, open-source intelligence identity-profiling system targeting Israeli governmental and non-governmental individuals, as well as Jewish diaspora organizations. In one workstream, the company said, the operator ran the program against a pre-existing target list. The operator also sought to generate open-source intelligence products about Israeli and U.S. people, Anthropic said.
• IRANIAN DOMESTIC SURVEILLANCE: Anthropic said it identified and banned 16 Claude accounts operated by two linked units associated with unidentified Iranian paramilitary and domestic security agencies. One of the organizations, which Anthropic described as a seven-department organization with offices across Iran’s provinces, claimed to maintain an identity-record database of Iranian nationals. The other used Claude to work on a malicious browser extension used to mass-harvest user identities from major social network platforms.
A request for comment sent to the Iranian government’s representatives at the United Nations in New York was not immediately returned.
• MALI COMMUNICATIONS SURVEILLANCE: Anthropic said Claude was used as the main engineering tool for a domestic surveillance platform built for Mali’s state intelligence service that could monitor data associated with about 25 million SIM cards across the country’s three mobile operators. The company said the system, called Lakana 360, appeared to have been developed largely by a single consultant working with Mali’s National State Security Agency.
A request for comment sent to the office of the prime minister of Mali was not immediately returned.
POLITICAL TARGETING, INFLUENCE OPERATIONS
CAMPAIGN AND PARTY TARGETING: Anthropic said it observed a French-speaking actor using Claude to target 42 European political parties, media, think-tanks and software-as-a-service providers used by those organizations.
The operator’s system stole data and login credentials for various services, and compiled a “purpose-built doxxing platform” that could be used to gather and expose various information on numerous people.
SCAMS
AI-DRIVEN DATING APPS: Anthropic said a China-based operation developed more than 20 dating apps containing over 4,700 AI personas that interacted with at least 25,000 users.
The operation combined automated personas with paid human workers to persuade users that the profiles were genuine, the company said.
(Reporting by Eduardo Baptista and AJ Vicens in Detroit, Editing by Rosalba O’Brien)
Comments