A Proposed Framework for Artificial Intelligence Safety and Technology Readiness Assessments for National Security Applications

Samuel Browne, Thomas Pike and Mark Bailey
Additional contact information
Mark Bailey: National Intelligence University

No ekth8, OSF Preprints from Center for Open Science

Abstract: The integration of artificial intelligence (AI) within the national security sector is increasing, yet traditional Technology Readiness Level (TRL) assessments fall short in evaluating the maturity of AI-enabled systems. Challenges specific to AI, such as alignment, explainability, and control, complicate the assessment of technical risk using the TRL framework. To address this gap, we introduce an AI Readiness Level (AIRL) framework, modified from TRL, to better evaluate the risk and readiness of AI systems in national security. The AIRL framework focuses on five key criteria: alignment, justified confidence, governance, human readiness level (HRL), and data readiness level (DRL). By assessing these dimensions, AIRL identifies minimum standards for deployment, areas for improvement, and technical risks. We apply the AIRL framework to hypothetical AI systems, revealing that despite high TRL ratings, these systems often fall short in critical AI-specific areas. AIRL offers program managers guidance on advancing AI systems to higher readiness levels, with a focus on improving alignment, transparency, governance, user interfaces, and data management. The framework provides valuable insights for policymakers, developers, and operators to ensure AI solutions meet ethical and performance standards in government operations.

Date: 2024-08-27
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
https://osf.io/download/66cceda4a66b610e2c1805da/

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:osf:osfxxx:ekth8

DOI: 10.31219/osf.io/ekth8

Access Statistics for this paper

More papers in OSF Preprints from Center for Open Science
Bibliographic data for series maintained by OSF ().