Undersea Weapons Specifications and Procurement Documents: Case GTG-17001

What happened

We identified a threat actor based in China who used Claude to advance three parallel workstreams for an anti-torpedo weapons system:

First, the actor used Claude to draft a Chinese-language specification for an anti-torpedo fire control system (the core logic for aiming and timing anti-torpedo weapon responses). The document was intended to obtain approval from a Chinese defense manufacturer, thereby advancing the work to technical certification and operational testing.

Second, the actor used Claude to produce a Chinese-language technical proposal of over 200 pages, accompanied by an executive briefing slide deck.

Third, the actor used Claude to benchmark its own system against specific US anti-torpedo and anti-submarine programs based on publicly available information. They then generated a Chinese-language briefing on US Navy systems based on open-source reporting.

The actor claimed to be an original equipment manufacturer in the US defense sector. We assess the actor to be linked to a Chinese defense industry manufacturer that aims to produce weapons specifications and procurement proposals for the People's Liberation Army Navy.

What AI did in this case

The actor used Claude to write the procurement proposal, refining it through multiple draft rounds. After each draft round, the actor instructed Claude to act as a hostile expert reviewer to critique the proposal, then used that feedback to improve the next version. At the same time, the actor used Claude to build parts of the anti-torpedo weapons system's fire control software and the test matrices to validate them.

The operational gain the actor achieved was using Claude to automate the production of complex technical outputs. The actor used the model to compress the development timeline for certification registration, compliance documentation, and automated fire control logic. The actor also accelerated the traditional human review cycle by having Claude play a role in multiple rounds of review to critique the procurement proposal.

What the report observed

The report confirms the existence of three parallel workstreams: a specification, a technical proposal, and a US program comparison briefing. The report confirms the actor claimed to be a US defense OEM but assesses it to be linked to China's defense industry.

Document generation does not equal actual weapons deployment. This site does not publish any technical specifications or fire control logic.

Confirmed & unknown

Platform response

Anthropic banned the relevant accounts and incorporated the investigation findings into its safeguards to reduce the risk of future abuse.

Limits of response：Banning accounts cannot prevent the actor from continuing document work using other tools.