Pillaging of the Digital Commons

How AI companies took everything public (and private) and sold it back to us

Photo by Mariia Shalabaieva on Unsplash

We didn’t know what we were doing on Facebook in 2007 when we publicly posted pictures of our dogs, our children, and our vacations and shared opinions about movies and music and politics. But we were actually training an AI model. We didn’t know because nobody told us. I’m sure that technically there was some line in the user agreement (which we were pressured to rapidly scroll through) which gave Meta permission to use our content for whatever they wanted to for any reason ever, but that’s the pathetic exchange we’ve all been making for decades now on the Internet. We get to do things for free and in exchange we give away our data and privacy and content. Facebook and then, in 2010, Instagram sold us on this vision of human connection, and we bought it. But in reality, the long game was that we were bought, our words and our images were bought, only for them to be sold back to us in the form of AI.

And it’s not just Meta. Elon Musk’s Grok is trained on X by default. You have to opt-out through a maze of setting menus to avoid training Grok. ChatGPT is trained on Reddit. Anthropic (Claude) was illegally trained on my book (Disruptive Witness) among many others. The AI music “generator” Suno scraped YouTube and other music sites to build its model. Over and over again, AI companies scraped massive amounts of data to build AI models, regardless of whose data it was.

Much of this data scraping has been of publicly available content, not copyrighted content, things said out in the open for anyone to read or listen to or watch: a recommendation I left on Reddit, a picture of my kid sliding down a slide on Instagram, a political rant on Facebook. I shared these things in public, so anyone could “pass by” and “read” them and “learn” from them. That’s all AI is doing, what any human could do.

But is it? Is the totalizing scrape of an AI machine for the purposes of creating an artificial intelligence which is sold back to me or used against me through ads actually the same thing as someone stumbling across my old political rants on Facebook from 2007? I don’t think it is. Not remotely. Instead, I think we have allowed AI giants to pillage the commons. The commons are the publicly owned spaces in our society. In this case, we have the digital commons, like Reddit. Where we meet and exchange ideas. Technically, these spaces are owned by corporations, but the content, the words and pictures should be ours. They should be publicly protected goods.

In allowing AI companies to scrape our commons, our content, our labor, our creations, we are doing their work which they then turn into profits. And probably the most nefarious part of all this is that 99.9% of us had no clue this is where we were headed when we started sharing content in 2007 or 2010 or whenever it was. We were sold one vision of social media and the Internet, and then that content was used for another purpose and sold back to us.

Even worse, a segment of the content that has been scraped is explicitly copyrighted. Thankfully there have been and will be some lawsuits, but those are mostly slaps on the wrists for multi-billion dollar companies.

The logic behind all this scraping is that it is better to move fast and break things (including laws and morality) in order to help humanity come into a better future. But there are two alarming problems with this logic. First, the intentional suspension of morality and law for the “good” of humanity is Nietzschean. It is the logic of the Overman. Morality and law are there to keep our worst instincts in check. If technology needs to move so fast that it must break morality and the law, then maybe technology is disordered. Second, it gives power to an elite few who control AI models to decide how to use those models and what is in the “best interest” of humanity. We should not tie humanity’s future to the vision of a few techno-optimists.

What should we do moving forward? If you are on Meta products, you can switch your accounts to “private” to protect future posts, get off Meta products, or choose not to post anything you would be uncomfortable training AI with (don’t post pictures of your children, etc). On Twitter you can switch off training. But in general, we should all keep in mind that what we post publicly will be used publicly, is not safe publicly. Platforms change. Technologies change. Who knows to what use our content will be put tomorrow?

Remember your friends and loved ones. Remember to meet with them away from phones and data and recording. They can’t scrape that.

---

A special thanks to the paid subscribers who came to the live Q/A on Monday. I’ll be hosting one of those quarterly as a perk to paid subscribers. You can watch the video here (if you are a paid subscriber!). Also note, if you are a Founding Member, I will happily sign and send you one of my books as a one-time perk. Just DM or email me your choice and address.

Finally, another special thanks to all those who took the time to fill out the reader survey. So many of you did that and left me kind notes. I read them all. And I am taking all this feedback to heart. Thank you.

-Alan