Fine-Tuning Unlocks What Alignment Was Hiding, Not What You Taught It

Sun, 21 Jun 2026 11:45:17 +0000

When researchers fine-tuned GPT-4o on a benign plot-expansion task — take a semantic summary, expand it into prose — it started reproducing verbatim passages from copyrighted books it was never shown during fine-tuning. Not just books by the authors it trained on. Books by completely unrelated authors. Cormac McCarthy unlocked by Haruki Murakami. Ta-Nehisi Coates spilling out from a Virginia Woolf fine-tune. Up to 460-word contiguous strings, reproduced at a rate climbing from under 8% baseline to over 91% after fine-tuning.

Copyright on BRYSGO

Fine-Tuning Unlocks What Alignment Was Hiding, Not What You Taught It