Treaty Proposals

Shorter-term proposals are important to understand insofar as they are useful for finding concrete policies to try to get decisionmakers to push for (i.e. what helps long-term that’s closer to the Overton Window?), and longer-term proposals insofar as they help with seeing how work fits into the ultimate goal of mitigating catastrophic and existential risks from AI. Many great policy/treaty ideas are bottlenecked by working out the exact details of their implementation in writing.

This page needs more details on the specific red lines currently being suggested by various groups.

Top external resources

• Shorter-term focused
  • “Global Call for AI Red Lines”
  • “In Support of Mandatory Nucleic Acid Synthesis Screening and Recordkeeping”
  • Mandating current best safety practices
  • Mandating transparency (I don’t have a resource for this yet, but worth mentioning)
• Longer-term focused
  • “AI 2040”. Proposal for a strong slowdown but not complete pause, and how this could be effectively enforced. Also includes proposals for mitigating extreme power concentration from AI.
  • “Narrow Path”. Pause until we can confidently make it safe
  • “An International Agreement to Prevent the Premature Creation of Artificial Superintelligence”. Pause until we can confidently make it safe

Parent pages: Main page