13,000-word essay argues organizational competence, not slowdowns, is the underrated key to AI safety
random_walker · x · 2026-09-15
Sayash Kapoor and Arvind Narayanan published a 13,000-word essay analyzing loss-of-control incidents at AI companies and what technical and policy interventions could improve safety. Derek Thompson amplified its core point:
- Organizational competence is a hugely underrated piece of AI safety. There's a frontier consensus to "pace" or pause, but if a company's RL still encourages misalignment or its sandboxes are poorly built, going 90mph or 60mph doesn't matter — operational incompetency is unsafe at any speed
- The polarization between cybersecurity and AI safety communities is counterproductive: the safety community frames incidents as an alignment crisis, while practitioners see them differently, hindering effective practice
The authors call for smartly designing, safely experimenting with, honestly sharing, and maybe eventually mandating security protocols for advanced AI.
More from Companies & People
- AI Twitter drama: users accuse Brian Chau of riding a 40x reach boost — anpaure · 2026-09-15
- Google named Fast Company's 2026 Design Company of the Year for AI-era design — GoogleDesign · 2026-09-15
- Microsoft researchers may threaten to quit over model welfare language, predicts artist — sterlingcrispin · 2026-09-15
- Hugging Face CEO, first public agent cyberattack victim, briefs DC policymakers — LysandreJik · 2026-09-15
- Why Jensen Accelerates While Dario Wants to Pause: A Theory of Misaligned Incentives — RachelVT42 · 2026-09-15
- China beats US in consumer AI adoption on super apps, Morgan Stanley survey finds — pstAsiatech · 2026-09-15