On Sept. 8th, Jacob Coxon announced his departure from his cushy job at Anthropic to warn us about what researchers already knew: artificial intelligence is dangerous. Coxon posted on X the details of his resignation, and how he felt that after working for both OpenAI and Anthropic—because only one was not enough— he finally understood how AI is not an executive functioning being and is in fact controlled by those who have the most power in society.
In all seriousness, I do believe that people are waking up to the realities of giving AI limitless access to human information and the real consequences of this. In Coxon’s initial post, he warned his readers to not underestimate the power of these tools. I’m sure we’ve seen multiple ways just this past year alone about how right he was.
In the world of surveillance, there is value in knowing the daily routines of the people you’re trying to track down. For obsessive ex-boyfriends, Flock Safety’s partnership with law enforcement made stalking that much easier. Nowadays, someone doesn’t even need to know the person they’re stalking to access direct information about them.
Aside from unwarranted surveillance, AI is also responsible for the thievery and destruction of intellectual property. 404 Media tracked an Amazon shipment of books and found out that AI companies were bulk buying classic books, destroying the bindings and scanning the text to train the language models. The number of books destroyed is unknown, but Anthropic planned to scan between 500,000 to two million books. That’s a staggering amount of information that no one machine should have, let alone to take away from real people.
Protests of AI only become viral once someone has already reaped the benefits of it. Quitting your job must seem like a brave thing to do, but that was only after accumulating the wealth that comes with working in the field of artificial intelligence. Coxon spent roughly three years pretraining these models at OpenAI and Anthropic.
If we as a society are going to amplify voices like Coxon’s, then we must start with those who fought against their comfort to bring attention to these issues. In Sept. 2025, Guido Reichstadter, Michaël Trazzi and Denys Sheremet.went on a hunger strike against Anthropic and GoogleMind to protest the ever growing issue of rapid AI development.
These were also men who worked for the very thing they are condemning. As brave as they were, it is also important to remember the real victims of AI.
Every day citizens whose neighborhoods are at risk of data centers being built, or who are facing the consequences of already built data centers. These consequences include power outages, water shortages and decreased quality of air. As edgy as a resignation post on the internet is, there are real life consequences to the hypotheticals that people are afraid of.
