How Do We Govern AI Well? — LessWrong
lesswrong.com · 9,380 words · saved by 1 readers
This is the first essay in my series How do we govern AI well? For more on the broader vision, see the introduction. …
x How Do We Govern AI Well? — LessWrong AI Frontpage 2 How Do We Govern AI Well? by kaime 19th Mar 2025 30 min read 0 2 This is the first essay in my series How do we govern AI well? For more on the broader vision, see the introduction. Abstract This essay argues that we face a fundamental problem in AI governance: governance mechanisms tend to degrade as AI capabilities advance . This degradation isn't random, it's systematic and predictable. The strongest evidence for this claim is that every governance approach we might consider (liability frameworks, auditing, industry self-regulation, etc
related reading
- Building Technology to Drive AI Governancebounded-regret.ghost.io
- The case for ensuring that powerful AIs are controlled — LessWronglesswrong.com
- The Case Against AI Control Research — LessWronglesswrong.com
- What failure looks like — LessWronglesswrong.com
- How do we (more) safely defer to AIs? — LessWronglesswrong.com
- Radical Optionality — Governing Transformative AI Under Uncertaintyradical-optionality.ai
- On Private Governancehyperdimensional.co
- AI Control: Improving Safety Despite Intentional Subversion — LessWronglesswrong.com
- What failure looks like — AI Alignment Forumalignmentforum.org
- AI policy must fail gracefullysubstack.com
- AI policy must fail gracefullyasteriskmag.substack.com
- The current bottleneck is political will, not research — LessWronglesswrong.com