Onlykashmir.in News Desk
Anthropic, the artificial intelligence company behind the Claude chatbot, has disclosed that a threat actor cell based in northern Yemen used its Claude Code tool to develop guidance software for weapons programmes, in a case the company has catalogued as GTG-87001 in its ongoing effort to detect and counter misuse of its AI systems.
According to the company’s account, the cell was running three separate weapons development programmes at the time it was disrupted. These included a guided rocket built around a commodity phone class flight computer with final phase homing guidance, a multi-stage ballistic missile with a stated range goal above 2,000 kilometres, and a multi-variant missile set referred to internally as “R2000,” which included a hypersonic glide vehicle variant, according to the disclosure.
Anthropic said the actors used Claude Code in place of human software engineers to build the guidance, navigation and control software that steers and stabilises a flying vehicle, including integrating an open source autopilot onto the flight computer, writing control and position estimation code, tuning control settings, running a firmware build pipeline and carrying out flight simulations. The company said the cell ran multiple Claude instances simultaneously, assigning each a distinct role in a workflow resembling a small engineering team, with one instance writing code, another conducting research and a third reviewing the first instance’s output.
The disclosure noted that Anthropic’s safeguards blocked many of the group’s requests, though not all of them, with the actors using tactics such as concealing their ultimate goals and the intended use of the software, and splitting their work across multiple sessions so that no single interaction revealed their full intent. Despite this, the company said it has no evidence the group succeeded in fielding an operational weapon, though the actors did test fire a guided rocket, a launch that appears to have failed, with the group returning to Claude within hours to try to diagnose the failure.
Anthropic said it identified the activity through its internal investigations into suspected weapons development, subsequently banning the accounts associated with the actors and sharing threat information with public and private sector partners to help mitigate the risk. The company cautioned, however, that the actors had already built an offline simulation toolkit independent of Claude or other engineering software such as MATLAB, suggesting the group retained some capability to continue its work outside the company’s platform.
