By Benson Daniel
Russian developers used Anthropic’s Claude artificial intelligence system to help develop software for autonomous “kamikaze” drones capable of identifying targets and carrying out attacks with limited human intervention, according to a new report by the AI company.
Anthropic said the activity was uncovered as part of an investigation into the misuse of its AI models between December 2025 and August 2026. The company identified several cases involving the use of AI for conventional weapons development, cyber operations, surveillance and other potentially harmful activities.
In the Russian case, developers reportedly used Claude to assist in creating software for swarms of attack drones. The system was designed to allow drones to analyse their surroundings, identify potential targets and coordinate their actions, raising concerns about the growing role of artificial intelligence in modern warfare.
The developers also used battlefield footage from Ukraine to support the development of machine-vision capabilities. Such technology can allow drones to interpret visual information and make decisions about objects or targets without requiring constant instructions from a human operator.
Anthropic said the Russian actors attempted to bypass restrictions placed on the use of its services in certain locations. The company subsequently disrupted the activity and took steps to strengthen its safeguards.
The revelations highlight the rapidly changing nature of the war in Ukraine, where both sides have increasingly relied on drones and other unmanned systems. Artificial intelligence is becoming an important part of that competition, particularly in areas such as target recognition, navigation, surveillance and battlefield coordination.
The development has also raised broader questions about the ability of technology companies to prevent advanced AI systems from being adapted for military purposes. Anthropic’s latest threat report identified cases involving actors from several countries and said increasingly sophisticated users were attempting to turn AI from a simple assistant into a tool capable of carrying out complex operations.
Another Russian-linked operation identified by Anthropic involved cyber espionage against Ukrainian government and military organisations. The group used AI to assist with activities including phishing, malware development and attempts to evade security systems.
The findings come as Russia and Ukraine continue to expand their use of unmanned weapons. Ukraine has also been developing increasingly sophisticated AI-assisted drones and other robotic systems, underscoring the emergence of a new form of warfare in which software, automation and artificial intelligence are becoming as important as conventional weapons.
The growing use of autonomous systems is likely to intensify debate over human control of weapons and the rules governing AI-enabled warfare. Governments and technology companies face increasing pressure to ensure that powerful AI tools cannot easily be repurposed to accelerate attacks or develop weapons capable of operating with little human oversight.
Anthropic said it has shared information about the identified activities with relevant authorities and industry partners as part of efforts to detect and prevent further misuse.
Leave a comment