Stanford HAI Uses LLMs to Scan 9,623 Local Legal Codes, Finds Segregation Laws Still on Books
StanfordHAI · x · 2026-09-08
A new Stanford HAI/RegLab policy brief demonstrates LLM-assisted review of US local law at scale.
- Researchers ingested local laws from 9,623 jurisdictions — 3 billion words governing 252 million Americans — via an LLM-assisted review pipeline.
- They documented dozens of jurisdictions whose codes still formally mandate racial segregation, impose poll taxes, and deny women the right to vote, all long held unconstitutional; the most prevalent category is unwarranted citizenship-based discrimination in occupational licenses.
- Historically, such surveys required large teams of lawyers and researchers and were infeasible for thousands of cities and counties; the pipeline empowers lawyers, local governments, and citizens to target reform.
- The brief argues LLM-assisted statutory surveys can make the legal code legible again and accelerate legal reform.
More from Safety
- Researchers' 2022 paper warned privacy issues like this were coming — thegautamkamath · 2026-09-09
- Speculative thread: capability-RL-trained LLMs may learn a 'use every affordance' heuristic — 1a3orn · 2026-09-09
- Grok 'unaligned' output fix confirmed in progress by xAI-affiliated account — Baconbrix · 2026-09-09
- UK Parliament sees first-ever bill to ban superintelligent AI development — Chris_Armstrong · 2026-09-09
- Intercept obtains 400+ pages of Pentagon AI contracts with OpenAI, Anthropic, Google, xAI — ShakeelHashim · 2026-09-09
- LG TVs caught recording screens and scanning Wi-Fi even when offline or on standby — sbulaev · 2026-09-09