Artificial intelligence company Anthropic stated in its threat intelligence report that Claude models are used in activities ranging from conventional weapons development to cyber operations, from military reconnaissance to political surveillance.
The report cited by Reuters included cases related to actors linked to China, Russia, Iran and Yemen, as well as operations targeting political organizations in Europe and the large-scale surveillance system developed for Mali.
While Anthropic stated that the relevant accounts were closed for some of the detected activities, it was pointed out that a significant part of the findings in the report were based on the company's own evaluations.
AIR DEFENSE SYSTEM WAS MODELED
According to Anthropic, a China-based actor used Claude to develop a software suite for electronic warfare and suppression of enemy air defenses.
At one stage of the project, 12 targets in Taiwan were added to the simulation. These included early warning radar, Patriot and Tien Kung missile batteries, air bases and a command bunker.
Anthropic claimed that the account information linked the actor to Chinese research institutions, including the Chinese People's Liberation Army Academy of Military Sciences. The company banned the accounts in question. The Chinese Ministry of Foreign Affairs stated that it was not aware of the report and stated that Beijing advocates the development of artificial intelligence for good purposes and opposes "distortions and smears" against China.
In another case in the report, it was alleged that a China-based actor used Claude to develop the technical specifications and fire control software of an anti-torpedo system for the Chinese navy.
It was noted that the actor prepared a technical proposal of more than 200 pages with artificial intelligence, compared the system with US Navy technologies and had simulated technical inspections to determine the shortcomings of the proposal.
Anthropic assessed that the actor is linked to a Chinese defense industry manufacturer that aims to develop weapons systems for the Chinese People's Liberation Army Navy.
It was also alleged that a China-linked defense-intelligence actor was using Claude to investigate foreign countries' high-power microwave weapons, their components, manufacturers and supply chains. According to Anthropic, the studies contained information that could help reverse engineer weapons and develop countermeasures.
'BALLISTIC MISSILE' STUDY FROM YEMEN
One of the most striking sections of the report was regarding Yemen.
Anthropic announced that a group in North Yemen used Claude in the development of a missile variant carrying guided rocket software, a ballistic missile planned to have a range of more than 2,000 kilometers, and a hypersonic glide vehicle.
It was stated that the actors used Claude for coding, simulation and troubleshooting technical problems, and also resorted to artificial intelligence after a guided rocket test that was considered unsuccessful.
Anthropic reported that there was no evidence that the group had successfully fielded an operational weapon, and that security systems blocked most but not all of the requests.
Reuters specifically noted that the Iran-backed Houthis control most of North Yemen, but Anthropic did not determine whether the actors in question were Houthis.
AUTONOMOUS FPV DRONE DRIVING
According to Anthropic, Russia-based actors, likely working independently, used Claude to develop software for a swarm of autonomous FPV attack drones.
Studies; It included terminal guidance, target selection and coordination of multiple aircraft. FPV drones are used extensively in both reconnaissance and attack missions in the Russia-Ukraine war.
In another case, it was stated that a Russia-based supply manager used Claude to use intermediaries in China and Hong Kong to purchase third-country routes that could disguise the final destination, to purchase products of European production and potential military use.
Products sought included German-made magnetometers, space-grade photovoltaic plates and aviation oxygen systems.
IDENTITY PROFILING AND INTERNAL SURVEILLANCE
Anthropic also claimed that an Iran-linked operator had developed an automated open source intelligence and identity profiling system targeting Israeli individuals and organizations and Jewish diaspora organizations.
It was also announced that 16 Claude accounts, allegedly operated by two units affiliated with paramilitary and internal security institutions in Iran, were closed.
It was noted that one of the units claimed to maintain an identity registration database of Iranian citizens, while the other was working on a malicious browser add-on to mass collect user identities from major social media platforms.
SYSTEM COVERING 25 MILLION SIM CARDS
It was also claimed that Claude was used as the basic engineering tool of the internal surveillance platform called Lakana 360, developed for Malian state intelligence.
According to Anthropic, the system was designed to be capable of monitoring data associated with approximately 25 million SIM cards across the country's three mobile operators and was largely developed by a single consultant working with Mali's National State Security Agency.
EUROPE IS ALSO THE TARGET
The report also included a case of political targeting activities.
It was stated that a French-speaking actor used Claude to target 42 political parties, media organizations, think tanks and the software services used by these institutions in Europe.
