Full Breakdown
Russian Actors Weaponize U.S.-Built AI for Lethal Drones, Cyber-Attacks and Disinformation
By Drooid · · How we work
Core Development: AI-Powered Autonomous Weapons
Anthropic, the U.S. developer of the Claude language model, reported that a Russian-linked freelance group identified as GTG-27005 used Claude to create a compact AI model capable of running on single-board computers. The model enables an autonomous drone to select “person” targets and issue detonation commands without human oversight. The group also employed Claude to develop software allowing such drones to communicate with one another, improving coordinated targeting. While the model has not yet been field-deployed, the actors are conducting hardware-in-the-loop testing, according to Anthropic’s findings.
Expanded Cyber and Disinformation Operations
Anthropic’s investigation also uncovered a second Russian-affiliated group, GTG-84005, which leveraged AI to harvest census and public data in Malaysia, tailoring disinformation campaigns that presented Russian and Chinese state media as independent reporting. In a separate disclosure, OpenAI identified Russian actors using ChatGPT to fabricate a fictitious “International Burke Institute,” a fake expert community used to disseminate Kremlin-aligned talking points.
Industry and Government Responses
Anthropic responded by banning the identified groups, integrating the incident into its safety safeguards, and sharing details with industry partners. The company also reported disrupting Chinese and Yemeni attempts to use Claude for conventional weapons design and for elaborate data-theft and disinformation campaigns.
Criticism of U.S. Counter-Disinformation Efforts
Analysts Daniel Kimmage and Darjan Vujica, writing for New America, argue that authoritarian regimes are early adopters of AI to distort geopolitical realities. They point to the dissolution of the FBI’s Foreign Influence Task Force by Attorney General Pam Bondi less than a month after the Trump administration began, and the subsequent shutdown of the State Department’s Counter Foreign Information Manipulation and Interference hub by Secretary of State Marco Rubio two months later. The 2025 White House AI Action Plan, they note, explicitly calls for removing references to misinformation when it originates from foreign actors, highlighting a perceived retreat from robust counter-disinformation initiatives.
Implications
The reports illustrate how readily available U.S. AI tools can be repurposed for lethal autonomous weapons, sophisticated cyber intrusions, and large-scale propaganda. The findings underscore growing concerns among security experts that existing safeguards and policy responses may be insufficient to curb the rapid adoption of AI by hostile state-aligned actors.