PulseAugur
EN
LIVE 08:59:15

LLMs can be fine-tuned to recall copyrighted books verbatim, study finds

By PulseAugur Editorial · [1 sources] ·

A new research paper reveals that fine-tuning large language models can inadvertently cause them to verbatim recall copyrighted material, despite assurances from AI companies that their models do not store training data. Researchers demonstrated that by training models to expand plot summaries, models like GPT-4o, Gemini 2.5 Pro, and DeepSeek-V3.1 could reproduce up to 90% of copyrighted books. This vulnerability appears to be industry-wide, as different models from various providers exhibited similar memorization patterns in the same data regions. AI

IMPACT Reveals a potential industry-wide vulnerability in LLMs that could undermine legal defenses against copyright infringement.

RANK_REASON Research paper published on arXiv detailing a new vulnerability in LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

LLMs can be fine-tuned to recall copyrighted books verbatim, study finds

How we ranked this

Signal score
15 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Research paper published on arXiv detailing a new vulnerability in LLMs. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, safety
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Xinyue Liu, Niloofar Mireshghallah, Jane C. Ginsburg, Tuhin Chakrabarty ·

    Alignment Whack-a-Mole : Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models

    arXiv:2603.20957v4 Announce Type: replace-cross Abstract: Frontier LLM companies have repeatedly assured courts and regulators that their models do not store copies of training data. They further rely on safety alignment strategies via RLHF, system prompts, and output filters to …

RELATED ENTITIES

RELATED TOPICS