A 300-word 100%-AI threshold: economist proposes a testable AI-content rule
paulnovosad · x · 2026-09-24
Economist Paul Novosad proposes a concrete rule for handling AI-generated content:
- Start by acting on the most egregious cases, like full pages at 100% AI.
- A fair threshold: any 300-word portion detected as 100% AI triggers action — but the rule should be tested against the pre-2022 corpus to ensure the false positive rate is low enough.
- He adds that permitted boundaries are imaginable, e.g. AI-generated exhibit notes with clear disclosure; but vague disclosures like "we used AI to refine sentences here and there" invite deception and abuse.
More from Safety
- AI Now: basic security protocols would have prevented the OpenAI/Hugging Face incident — AINowInstitute · 2026-09-24
- Guardian: AI Overviews reshape search as media outlets sue Google over lost traffic — nordicinst · 2026-09-24
- Dario x-risk-pills the UN Security Council while touting Anthropic's safety record — teortaxesTex · 2026-09-24
- Using eBPF to contain misbehaving AI agents: kernel-level network sandboxing — sloppenheimer · 2026-09-24
- Krueger's bill breakdown: federal AI department, superintelligence ban, 10^25 FLOP pause — DavidSKrueger · 2026-09-24
- Critical WordPress RCE CVE-2026-87902 reproduced for $4.93 with an AI agent — evilsocket · 2026-09-24