Houthis Used Claude Code to Develop Missile Guidance Software: Anthropic · Clash ReportSkip to contentLive Intelligence— — · --:--SearchDark modeHomeWorldAfricaAmericaArcticAsia-PacificEuropeIsrael-Gaza WarMiddle EastUkraine - Russia WarAll World news →DefenseFocusOpinionGalleryInfographicsPodcastHome/World/Africa/Houthis Used Claude Code to Develop Missile Guidance Software: AnthropicWorldHouthis Used Claude Code to Develop Missile Guidance Software: AnthropicAnthropic says operators in northern Yemen ran parallel Claude Code instances to design guidance software, simulate trajectories, and analyze a failed rocket test, effectively compressing specialist missile-engineering work into an AI-assisted workflow.ShareSaveText SizeA−AA+18pxA−AA+18pxClash ReportYemen's Houthi rebels launching a ballistic missile in Sanaa, March 27, 2018 - AFPAhmet Koçak11 Sept 2026 · 12:40 GMT · 3 MIN READGoogle SearchChoose Clash Report as a preferred source and see our reporting first in Top Stories.Add as preferred sourceA cell in northern Yemen used Anthropic’s Claude Code to perform work that would ordinarily require a missile engineering team, developing software for guided rockets, a long-range ballistic missile and a hypersonic glide vehicle concept.Anthropic’s September threat report describes operators running several Claude instances at the same time, dividing tasks between coding, research and technical review.The company assessed the cell as highly likely to be linked to the Houthis, according to the material provided.The operation marks one of the clearest examples in the report of generative AI being applied directly to conventional weapons development rather than used simply for research or information gathering.Parallel AI Agents Replace Specialist WorkThe Yemen-based operators used Claude Code across multiple engineering functions.Their projects included guidance software for a tactical guided rocket, a ballistic missile with a range exceeding 2,000 km, and a hypersonic glide vehicle variant called “R2000.”Rather than assigning the work sequentially to a single model instance, the group operated several sessions in parallel.One could produce code while another conducted research and a third reviewed the output.The workflow let a small group reproduce several functions normally distributed across specialist engineering teams.From Flight Computers to Trajectory SimulationThe technical work extended into navigation, flight control and simulation.The operators sought to integrate open-source autopilot software with a phone-class flight computer and used Claude to write navigation and control code.They also developed six-degree-of-freedom trajectory simulations, a method for modeling an object’s movement and rotation through space.They also used Claude for reinforcement learning work to tune flight-control algorithms.The operators ultimately compiled the project into a standalone executable that could run offline. That meant development could continue without direct access to Claude.Rocket Test Followed by AI Failure AnalysisThe activity moved beyond software development.The group test-fired a guided rocket in Yemen, according to the material documented by Anthropic. The test appears to have failed.Within hours, however, the operators returned to Claude and began analyzing launch telemetry to investigate the failure.That sequence, software development, physical testing and rapid post-test analysis, illustrates how the model was incorporated into an iterative weapons-engineering cycle rather than being used for isolated technical questions.Anthropic said it found no evidence that the group succeeded in fielding an operational weapon.But by the time the accounts were disrupted, the operators had already assembled an offline engineering toolkit that no longer depended on access to Claude.Safeguards Met With EvasionAnthropic’s safeguards blocked numerous requests during the project.The operators adapted by obscuring the intended end use of individual tasks, dividing work across separate conversations, and using other methods to circumvent restrictions.Anthropic subsequently banned accounts linked to the activity.The case highlights a problem for AI safety systems: individual requests may appear disconnected from weapons development when a larger engineering project is deliberately fragmented across sessions.Six Conventional-Weapons CasesThe Yemen operation was one of six conventional-weapons cases Anthropic documented.Three were linked to China, two to Russia and one to Yemen. The cases covered attempts involving missiles, armed drones, firearms and bombs.The wider report covers operations Anthropic said it disrupted between December 2025 and August 2026 across cyber operations, influence activity, surveillance, conventional weapons, biological misuse, scams and fraud, and illicit model distillation.Anthropic also described biological research cases in which it could not establish whether scientists were conducting legitimate research or pursuing weapons-related objectives.“You are not seeing someone in a comic book kind of way say, ‘Hey, I want to build a biological weapon to kill everybody,’” Jacob Klein, Anthropic’s head of threat intelligence, said. “It’s an incredibly nuanced situation.”For the Yemen case, the immediate outcome was more concrete: Claude had been incorporated into a real development cycle spanning design, simulation, testing, and failure analysis.The weapon itself was not shown to have become operational. The engineering capability assembled around it had already begun moving offline.About the AuthorAhmet KoçakClash ReportAhmet Koçak is a news editor at Clash Report based in Istanbul. He previously served as Deputy Managing Editor at Türkiye Today, helping launch the digital news platform in 2023, and spent three years as Senior Editor at Daily Sabah. His work focuses on breaking news, geopolitics, international affairs, and digital journalism.All articles by Ahmet KoçakGoogle SearchChoose Clash Report as a preferred source and see our reporting first in Top Stories.Add as preferred sourceSourcesClash ReportTagsAnthropicClaudeClaude CodeYemenHouthisballistic missilesAImissile guidanceweapons developmentdefense technology← Previous StoryRussian Drone Hits Route of Train Carrying Top EU AdvisersNext Story →Kremlin Recruited Undercover German JournalistRelated StoriesMore on this storyWorldRussian Drone Hits Route of Train Carrying Top EU Advisers13 SeptWorldKremlin Recruited Undercover German Journalist13 SeptWorldDemocrats Eye Sweeping Trump Probes If They Win Midterms13 SeptWorldScottish, Welsh, and Irish Leaders to Demand UK Independence13 SeptWorldUK Police Disregarded Andrew-Epstein Evidence: Whistleblower13 SeptWorldCanada Explores EU Membership After US Trade War13 SeptWorldEU Pledges €6.1B More for Ukraine Defense11 SeptWorldRussia Seeks to Influence German, French Elections: Sweden11 SeptWorldDefenseFocusOpinionGalleryInfographicsPodcastWritersFrom the world's conflict zones impartial and data-driven journalism.NDownload on theApp StoreGet it onGoogle Play© 2026 Clash Report. All rights reserved.About UsContactPrivacyTermsCookiesWhen the front line moves overnightEP 1470:000:001.0×Now playing · ClashPointMinimizeEP 147When the front line moves overnightGuest: Daniel Reyes0:000:001.0×

Operators in northern Yemen utilized Anthropic’s Claude Code to execute tasks traditionally requiring a specialized missile engineering team, effectively streamlining the development of software for guided rockets, a long-range ballistic missile, and a hypersonic glide vehicle concept. This operation involved running multiple Claude instances simultaneously to divide the workload among coding, research, and technical review, demonstrating a workflow that compressed complex engineering tasks into an AI-assisted process. The projects encompassed developing guidance software for a tactical rocket, a ballistic missile with a range exceeding two thousand kilometers, and a variant of a hypersonic glide vehicle referred to as “R2000.” The collective effort integrated navigation, flight control, and simulation, allowing the operators to integrate open-source autopilot software with a phone-class flight computer, using Claude to generate navigation and control code. Furthermore, the group developed six-degree-of-freedom trajectory simulations and employed Claude for reinforcement learning to fine-tune flight-control algorithms. They successfully compiled these efforts into a standalone executable, establishing an offline engineering toolkit that could function without continuous access to the AI model.

The process extended beyond mere software development into physical testing and iterative analysis. The group tested a guided rocket in Yemen, which subsequently failed. Following this test, the operators returned to Claude to analyze the launch telemetry and investigate the failure, illustrating how the model was integrated into an iterative weapons-engineering cycle rather than being used for isolated technical inquiries. While Anthropic found no evidence that the group succeeded in fielding an operational weapon, the activity showcased how AI was incorporated directly into the entire cycle of design, simulation, testing, and post-test failure analysis. The operators sought to evade Anthropic’s safeguards by fragmenting requests across separate conversations, obscuring the intended end use of individual tasks, and employing other methods to circumvent restrictions, which consequently led to the banning of associated accounts by Anthropic.

This incident is documented as one of six conventional-weapons cases reported by Anthropic, which also involved operations linked to China, Russia, and Yemen, covering the development of missiles, armed drones, firearms, and bombs. Anthropic's broader threat report also addressed biological research cases, emphasizing that the context surrounding scientific work is crucial when assessing intent. The case highlights a potential vulnerability in AI safety systems: that individual requests, when intentionally fragmented across sessions for a larger engineering goal, may appear disconnected from the development of weapons systems. Ultimately, the Yemen operation demonstrated that advanced AI models can be incorporated into real-world, operational engineering cycles, moving the assembled engineering capability into an offline format even in the face of safety countermeasures.