AI research and discussion on LessWrong. www.lesswrong.com
ControlAI proposes three stopgap policies against immediate superintelligence-related risks: securing weapons-grade AI against theft, criminal liability for AI containment leaks, and government kill-switches.
Speculative commentary arguing that a US government 'national security' tier of frontier AI, with pre-release vetting and public-private partnerships, is emerging after Claude Mythos's hacking capabilities, and naming likely political players.
Argues that open-weight models undermine the defenses proposed for near-future AI risks, since guardrails can be fine-tuned away, use cannot be monitored, and bad actors have access by default.
Commentary on a RAND report claiming a superintelligence could be contained by hardening world infrastructure against harm, arguing the proposed safeguards are weak and implementation and buy-in remain major obstacles.
Essay arguing that falling AI development costs and easy model theft will lead to proliferation of offensive capabilities, drawing analogies to cryptography and estimating training cost trends for dangerous AI.
Analyses whether intent-aligned defensive AI can counter takeover-level adversarial AI, proposing threat models, a three-pillar defense framework, and legal constraints, and concludes the offense-defense balance remains uncertain.