ABC News
Details
- Date Published
- 4 Aug 2024
- Priority Score
- 2
- Australian
- Yes
- Created
- 4 Aug 2026, 10:00 am
Authors (1)
- Melanie TaitNEW
Description
Amid rampant reports of AI companies buying up masses of rare and out-of-print books to train large language models, and Australian second-hand booksellers say they are also fielding some curious enquiries.
Summary
This report examines the practice of AI companies acquiring massive datasets of rare and out-of-print Australian books to train large language models. While the segment highlights data sovereignty and intellectual property concerns, it touches on the broader issue of how frontier AI capabilities are built upon uncurated or proprietary cultural data. The acquisition of such specialized datasets contributes to the increasing power of LLMs, though the discussion focuses more on copyright and industry ethics than on direct catastrophic or existential safety risks.