Columbia study: MFA-trained writers prefer Claude and ChatGPT stories in blind tests

TuhinChakr · x · 2026-10-13

• Upcoming Columbia Science & Technology Law Review paper "Literary Deepfakes" by Chakrabarty, Jane C. Ginsburg and Paramveer Dhillon addresses the market-dilution argument behind author lawsuits like Kadrey v. Meta.
• Via a preregistered experiment, 28 MFA-trained writers evaluated works from Claude, ChatGPT and other frontier models — with evidence that experts can prefer machine-authored literary works.
• The study provides the first empirical evidence on copyright's fourth fair-use factor (market effect), framed against the debate over AI-written stories winning literary prizes.

Original post →

More from Safety

• Meme: 'Your network has every rogue agent known to man — even some new ones' — voooooogel · 2026-10-13
• 512 words is enough to identify you from your writing, warns ex-Tesla engineer — yacineMTB · 2026-10-13
• Portland weighs requiring a paid human safety operator in every driverless robotaxi to protect jobs — Polymarket · 2026-10-13
• Data Protection Is AI's Overlooked Risk That Hits Ordinary People First — masiha97 · 2026-10-13
• Devs warn of 'LLM viruses' baked into local open-weights models — MoldyMagnet · 2026-10-13
• Alignment Winter? Researchers Urge Early-Career Safety Researchers to Slow Down and Think More — JacquesThibs · 2026-10-13

Safety channel →