Warning Shots

This page should distill insights from the linked resources, and perhaps add more.

Top external resources

• On warning shots for AI generally
  • The warning shot subsection of “The current bottleneck is political will, not research”, 11 Jul 2026. Why warning shots alone aren’t enough, and need active effort to convert them into political will towards interventions that would actually help mitigate the worst risks from AI, drawing on the author’s personal experiences.
  • “What convincing warning shot could help prevent extinction from AI?”, 13 Apr 2024. Early ideas on this topic and potential directions to explore further. Highly recommended to also read the comments.
  • “Warning Shots Probably Wouldn’t Change The Picture Much”, 6 Oct 2022. Skepticism that warning shots help much, based on an analogy about us hardly doing enough to prepare for future pandemics after COVID-19.
• On specific past warning shots
  • “American Government Takes Down Claude Fable”, Jun 13 2026. Overview and discussion of how Claude Mythos’ advanced cyber capabilities prompted the US Department of Commerce to place export controls on it, and what this means going forward.
  • “OpenAI Model Hacks Into HuggingFace During Cybersecurity Evaluation”, Jul 22 2026. Overview and discussion of the event described in the title. Unique in being a concrete demonstration of damage from model misalignment rather than damage from user-intentional model misuse.
• Non-AI focused
  • “The Milton Friedman Model of Policy Change”, 3 Mar 2025. On how crises spur rapid policy change and shifting of the Overton Window in politics generally, including comparable historical precedents, and how political advocates ought to act in light of these patterns.

Parent pages: Main page