Army moving out on AI BOM concept for Project Linchpin

WASHINGTON — The Army is asking industry how to address potential vulnerabilities in the AI supply chain for Project Linchpin, the service’s first program of record to help build out a trusted artificial intelligence and machine learning pipeline. 

Through Project Linchpin, the service wants to develop an AI/ML operations “environment” for intelligence, cyber and electronic warfare systems. The program is run out of the Army’s Program Executive Office for Intelligence, Electronic Warfare and Sensors, which released two new requests for information on Oct. 30 for an AI “bill of materials,” or BOM, and “computer vision capabilities.”

“Project Linchpin is an initiative to enable the Army to establish necessary environments and services for AI-enabled systems, providing a mechanism to create new industry partnerships, and fostering a competitive environment for third-party integration into Army modernization programs within PEO IEW&S,” according to the AI BOM RFI. “The goal is to create a complete and efficient AI and ML development and delivery operational pipeline (AI/MLOps) with supporting services for sensor programs within PEO IEW&S while managing cost, schedule, risk and performance.”

The AI BOM, an idea floated earlier this year by Young Bang, principal deputy assistant secretary of the Army for acquisition, logistics and technology, would be structured like a software BOM, or S-BOM, used to secure the software supply chain. In August, Bang said AI BOMs will be critical for the Army when it comes to “catching up” to China in the AI race. 

According to the RFI, the AI BOM will consist of a minimum of three components: an S-BOM that will contain “the details and supply chain relationships of the components used to build and validate the AI model” details about the model, including training data, and data lineage used to create the model. The Army is asking industry for information regarding the cost of producing an AI BOM, a list of tools and processes required to implement an AI BOM into an AI/MLOps pipeline, the effectiveness of an AI BOM and any alternatives that could address vulnerabilities in the AI supply chain.  

Meanwhile, the service will be collaborating with the Pentagon’s Chief Digital and AI Office for computer vision capabilities “to detect and classify objects” using two specific controlled unclassified information labeled datasets, according to the RFI: horizontal full motion video from sensors mounted on ground combat vehicles and overhead imagery from satellites.

The Army is asking industry for feedback on their approach to training models, what skillsets and team structure is required to deliver trained models and what the cost would be to deliver trained models and their intellectual property strategies, according to the RFI. 

Industry has until Dec. 1 to respond to both RFIs. 

In September, the service awarded Booz Allen Hamilton and Red Hat a contract with a combined value of up to $2 million to support research efforts for Project Linchpin. The companies would specifically work on the Traceability, Observability, Replaceability and Consumability framework, which “seeks to ensure model and data integrity, data openness, and modular open system architecture (MOSA) design,” according to the announcement. 

“This contract is a significant first step to decouple AI from software, decompose components within a MLOPs pipeline, and wrap layers of security around the entire process,” Col. Chris Anderson, project manager for intelligence systems and analytics, said then. “These design principles will allow the Army to leverage the best of breed technology available across industry, academia, and government.”