The Getty makes nearly 88k art images free to use

AI training and open-licensed images

• Several comments argue AI models should be trained on openly licensed / public-domain images instead of scraped copyrighted art.
• Suggested mechanisms: social platforms adding license options on upload, public campaigns to normalize open licensing.
• Benefits mentioned: consent-based datasets, better public trust, more equitable access to training data, and pressure to clarify licenses.

Artist incentives and concerns

• Some question why artists would license work for AI, noting broad hostility to AI art and job fears.
• Others liken it to open source: done for ideology, reputation, or donations, not direct profit.
• A recurring point: artists want control more than money; revenue-sharing schemes don’t address that.
• There’s disagreement over whether an individual artist “loses little” by allowing their work into training data.

Scale and sample efficiency

• Critics say public/open collections are far too small for current model scales (e.g., 88k vs billions of images).
• Proponents reply this is exactly why research into sample-efficient models would be valuable.

Copyright, public domain, and reproductions

• Strong debate over whether photos/scans of public-domain art are themselves copyrighted.
• U.S. case law cited (Bridgeman) and Copyright Office guidance are used to argue that exact “slavish” reproductions are not copyrightable.
• Others emphasize practical realities: publishers demand permissions; museums control physical access and can contractually restrict use even if copyright doesn’t apply.
• Jurisdiction differences noted (e.g., recent UK ruling limiting museums’ reproduction fees).

Value of the Getty release

• Some see it as mostly symbolic because much of the art is already public domain.
• Others stress the real value is high-quality digitization, hosting, and metadata curation, which are labor- and cost-intensive.
• Observations that public libraries and other GLAM institutions already do similar work, though often with fewer resources or poorer UX (low resolution, watermarks, broken links).

Technical and practical reactions

• Some users report broken or incomplete downloads for the highest-resolution files, suspecting load or server limits.
• People are interested in using the images for personal projects, bots, color analysis, and AI training, though 88k images are considered tiny compared to existing AI datasets.

Perceptions of Getty and its legacy

• Clarification that this is the J. Paul Getty Museum, not Getty Images, which has a reputation for aggressively licensing even public-domain material.
• Mixed views on the founder’s legacy: some emphasize the generosity of the endowment and current open access; others criticize his personal behavior and motives.
• Separate praise for the museum as a physical place: architecture, views, and special exhibits are highlighted positively.