The Asilomar AI Principles
Participants of the 1986 Asilomar conference on future technologies · 1986 CE
Excerpt
“Research in advanced AI should proceed only where there is a very high probability of benefit and a very low probability of harm. (paraphrase of the 1986 principles)”
Why the atlas keeps it
Drafted over a few hours on the final day of a four-day 1986 conference on future technologies, and the first written set of research principles for advanced AI agreed by the researchers themselves. Its structure — general principles plus separately stated capability thresholds — is the structure of every frontier safety framework since. The excerpt below paraphrases the principles rather than reproducing the agreed wording.
Terms this text reaches for
Ranked by the codex search index against the summary above, not hand-picked.
Frontier safety framework
governanceA developer's published promise of what it will do at each capability threshold.
AI safety case
governanceA structured, auditable argument that a system's residual risk is acceptable.
AI safety
alignmentThe practice of reducing the chance that AI systems cause harm, however they are built.
Capability threshold
governanceA pre-specified level of capability that triggers an obligation or a prohibition.
Frontier developer
governanceAn organisation training systems at the edge of available capability.
Constitutional AI
technicalTraining a model to critique and revise its own outputs against a written set of principles.