The Asilomar AI Principles are a set of 23 guidelines for the development of beneficial artificial intelligence, produced at the Beneficial AI 2017 conference convened by the Future of Life Institute at the Asilomar conference grounds in California in January 2017. The principles were published by FLI on August 11, 2017. The name and venue deliberately echo the 1975 Asilomar Conference on Recombinant DNA, at which biologists agreed on safety guidelines for genetic engineering.
The 23 principles are grouped into three categories. Five concern research issues, including the recommendation that the goal of AI research should be to create beneficial intelligence rather than undirected intelligence. Twelve concern ethics and values, covering safety, transparency, human control, privacy, and the avoidance of an arms race in lethal autonomous weapons. Six concern longer-term issues, including handling the risks posed by advanced AI and the principle that superintelligence, if developed, should be used for the benefit of all humanity. One widely cited principle states that AI systems should be safe and secure throughout their operational lifetime, and verifiably so where applicable and feasible.
The principles were drafted through discussion and voting among the conference attendees and then opened for endorsement. They attracted thousands of signatories, including a large number of AI and robotics researchers alongside other supporters. They carry no legal force, functioning instead as a voluntary statement of shared values.
The Asilomar Principles became one of the most frequently referenced early frameworks in AI governance, cited in subsequent policy documents and corporate AI principles. They sit alongside FLI’s 2015 research-priorities letter and its later 2023 pause letter as part of a recurring effort to shape norms around advanced AI.