Revenue-tied AI safety standard could shift industry behavior by late 2027, argues new regulatory proposal
ronbodkin · x · 2026-11-18
Former congressman @sliccardo proposes a novel AI regulation approach: since traditional legislation alone can't make AI safe, regulators should tie the standard of care to lab revenues—what AI labs care most about—to incentivize rapid development of safety technology.
Ron Bodkin's commentary:
- Tying safety standards to customer revenue is new; if PSS criteria are published by mid-2027 with a short qualification deadline, behavior could change by late 2027.
- Technical evals are necessary but not sufficient (sandbagging); criteria should also cover development process—autonomy risks from automated AI R&D and the developer's own oversight.
- Cites @dgrobinson, who wrote safety reports for 12 OpenAI launches, warning on departure that companies "aren't being nearly careful enough"—the problem is culture.
- Encouraged by reports that Trahan plans to revise the FRONTIER bill into a federal standard.
More from AGI Musings
- 'This is how recruiters, job-seekers and hiring managers work now too' — adam_dorr · 2026-11-18
- Former SEC Chair Gary Gensler says the AI bubble will likely burst and cause a recession before 2028 — SumitGup · 2026-11-18
- AGI hype is driven by grift and attention, argues Reddit skeptic — Present-Chocolate-24 · 2026-11-18
- AI Interviewers Are Turning Hiring Into an Enterprise AI Failure, Analyst Warns — DavidLinthicum · 2026-11-18
- Kolmogorov complexity, AIXI and undecidability: why intelligence lives in problems compute can't swallow — JensHonack · 2026-11-18
- BAIO101: An illustrated guide of 101 AI x bio concepts, from protein design to virtual cells — DrMorganLevine · 2026-11-18