I am one among the many (artists, musicians, writers, developers), who continue to be appaled at tech billionaires profiting from this [2]. As such, I have mostly ceased publishing original content online, while I wait for effective defenses to emerge.
The two industry-standards are `robots.txt` and captchas. Neither works: the former is simply ignored, even by industry heavyweights [3]; the latter is trivial to bypass (using AI itself [4], remote solving services [5], or anti-detect browsers/impersonators [6] + residential proxies [7]).
Emerging strategies include special fonts for creating OCR-resistant documents, tho current options don't seem to be effective [8]. Under the "dark forest" theory [9], paywalling or private distribution seem viable options, provided you can trust the recipients to not leak the content; of course, this is a death sentence for the libre ethos of the past 50 years, tho perhaps it was long overdue [10].
Please share your thoughts on any alternatives, real or speculative.
[0] https://news.ycombinator.com/item?id=49752056 - Microsoft exec called AI scraping 'the largest theft of labor in human history' (techcrunch.com)
[1] https://news.ycombinator.com/item?id=41275073 - Ex-Google CEO: AI startups can steal IP and hire lawyers to 'clean up the mess' (theverge.com)
[2] https://news.ycombinator.com/item?id=40297027 - StackOverflow is banning accounts that delete answers in protest against OpenAI (benui.ca)
[3] https://news.ycombinator.com/item?id=50401213 - Brave Sells Its Search Index to AI, and Your Website Is Probably in It (searchenginejournal.com)
[4] https://news.ycombinator.com/item?id=49930138 - Meta's Muse is fantastic for web scraping (sigh.dev)
[5] https://captchathecat.com/
[6] https://news.ycombinator.com/item?id=42547820 - Curl-Impersonate (github.com/lexiforest)
[7] https://news.ycombinator.com/item?id=48864252 - An update on residential proxies and the scraper situation (lwn.net)
[8] https://news.ycombinator.com/item?id=48873513 - Ghost Font: A font that humans can read but AI cannot (mixfont.com)
[9] https://news.ycombinator.com/item?id=47566442 - The Cognitive Dark Forest (ryelang.org)
[10] https://news.ycombinator.com/item?id=50224074 - PHP Foundation 'under-funded' with one full-time staff member, director says (reuters.com)
[–] Terr_ 10 hours ago
For example, a right to all their model-weights on request, plus permission to republish them.
Granted, this relies on some long-shot ideas, and the assumption that the legal system won't just say "big companies can auto-hack their way around a contract, but regular people have to obey it."
Invent a conlang, and then communicate only in that.
Franchise your work, so the universe has multiple authors. Let your audience also be creators.
Encrypt stuff and send out license keys. More friction, more adversarial cost.
Use a non-English Unicode programming language, where content (data) is in a binary.
Mess with formats. A static image is slightly tweaked and served base64-encoded. A video just shows the same thing for hours.
Distribute works on floppy disks, or only offer to fax the https:// link.
Create something that is an amalgamation of various Library of Babel artifacts, so the collage is merely a novel derivative.