Saab and MIT Partner with DARPA on ASIMOV Programme to Develop Ethical AI Benchmarks for Autonomous Systems
Saab, Inc. and the Massachusetts Institute of Technology (MIT) have joined forces under a contract awarded by the US Defense Advanced Research Projects Agency (DARPA) for the ASIMOV (Autonomy Standards and Ideals with Military Operational Values) programme. This initiative aims to establish benchmarks to objectively measure the ethical readiness of autonomous systems for military use.
The ASIMOV programme addresses the rising prevalence of artificial intelligence (AI) and autonomous technologies in both military and civilian domains. It seeks to create an automated framework for assessing how well autonomous systems perform ethically within military operational values, ensuring these technologies align with ethical principles in dynamic operational environments.
To meet ASIMOV’s objectives, Saab and MIT are developing the RESPECT (Responsible Ethical Standards for Pioneering Effective Combat Technologies) framework. This effort is led by Dr. Devesh Upadhyay and Dr. Christopher Vo from Saab and Professors Chuchu Fan and Nicholas Roy from MIT. Together, the team combines expertise in AI safety, formal methods, verification frameworks, and operational autonomy across diverse applications.
The RESPECT framework will utilise a generative modelling environment guided by large language models (LLMs) to simulate ethically challenging scenarios. This iterative approach will allow the benchmarking of ethical sensitivities, enabling autonomous systems to adapt and make decisions that are both operationally efficient and ethically sound.