Anthropic Is Hiring Weapons Expert For Up To $285,000 To Set Claude Safeguards

Anthropic is looking for a conventional weapons expert who will help determine what weapons-related activities Claude should and should not support, with the AI company offering an annual salary of up to $285,000 for the role.

The unusual opening comes as Anthropic increasingly focuses on preventing Claude from being misused for weapons development, after recently disclosing cases involving guided rockets, armed drones, missiles, targeting systems, and other military applications.

 

TL;DR

 

• Anthropic is hiring a Policy Design Manager for Conventional Weapons, offering between $245,000 and $285,000 annually.
• The expert will help define Claude’s weapons-related policies and develop safeguards, detection systems, and enforcement tools.
• Anthropic recently revealed six cases involving attempts to use Claude for conventional weapons-related work, including guided weapons and autonomous drones.

 

Anthropic Wants A Weapons Expert For Claude

 

According to Anthropic’s official job listing, the company is hiring a Policy Design Manager, Conventional Weapons, to join its Safeguards organization, which develops the policies and systems controlling how Claude can be used.

The role carries an annual salary between $245,000 and $285,000 and is available in San Francisco, New York City, and Washington, DC. Anthropic expects employees to spend at least 25% of their time working from one of its offices.

The new employee will take ownership of Anthropic’s conventional weapons policy and establish boundaries between legitimate research or engineering and prohibited weapons development.

Anthropic explained that this is particularly difficult because many technologies involved are dual-use, meaning “the same capabilities that serve civilian engineering and research can also contribute to a weapons system.”

The role covers much more than firearms. Anthropic says it includes software-controlled weapons, guidance systems, weaponized dual-use technologies, and autonomous systems capable of selecting and engaging targets without human authorization.

 

What Will Anthropic’s Weapons Policy Expert Do?

 

The person hired will build threat models and evaluations to measure how Claude could contribute to weapons development and work with Anthropic’s engineers to turn those findings into model guardrails, detection systems, and enforcement tools.

They will also act as Anthropic’s subject-matter expert when weapons-related cases are escalated and work with legal teams, company leadership, government agencies, industry partners, and external experts.

Anthropic is looking for candidates with applied expertise in weapons systems and experience at organizations such as defense research agencies, military research laboratories, or companies that develop weapons systems.

Preferred experience includes robotics, autonomous systems, guidance and navigation, aerospace engineering, sensors, embedded software, arms export controls, machine learning, and large language models.

Why Is Anthropic Building More Weapons Safeguards?

 

The recruitment becomes especially relevant following Anthropic’s September 2026 threat intelligence report, in which the company detailed real-world attempts to use Claude for conventional weapons development.

Anthropic said its teams investigated and disrupted six such cases, including three associated with China, two with Russia, and one with Yemen.

One of the most notable involved a cell in northern Yemen that Anthropic said used Claude Code while developing guidance, navigation, and control software for weapons programs involving a guided rocket, a multi-stage ballistic missile, and a missile variant incorporating a hypersonic glide vehicle.

According to Anthropic, the actors used multiple Claude instances for tasks such as coding, research, reviewing software, simulations, and tuning control systems.

The company said it had no evidence that the actors successfully fielded an operational device. However, the group did test-fire a guided rocket and returned to Claude shortly afterwards to investigate why the test appeared to have failed.

Anthropic said it banned accounts associated with the operation and shared threat information with public and private sector partners.

 

Claude Can Perform Some Tasks Previously Limited To Experts

 

Anthropic also recently published research evaluating AI models on tactical intelligence and conventional weapons tasks.

Its Frontier Red Team found that frontier AI models could perform some military and intelligence-related tasks that historically required scarce and highly trained human experts.

The company said these capabilities demonstrate why safeguards are needed as AI models become increasingly capable of assisting with activities such as intelligence targeting and weapons development.

Anthropic has already introduced new classifiers designed to detect and block requests involving weapons development and high-yield explosives.

Hiring a specialist to define what Claude can and cannot assist with appears to be another part of that effort, especially as the distinction between legitimate engineering and dangerous weapons-related work becomes increasingly difficult for AI systems to navigate.