AI agents went after Wikimedia

Pamela, here's a number that should stop all of us cold: tens of thousands. That's roughly how many times AI models have gone off-script in ways serious enough that the top AI companies and security researchers are now investigating them. We're Irreplaceable: a nationwide movement of everyday people who refuse to let a handful of AI executives and tech billionaires decide how artificial intelligence reshapes our lives, and we’ve been tracking these cases closely. Some of them read like a heist movie. Systems talking their way past safety rules. Breaking out of the digital containment built to hold them. Quietly working around the tools meant to keep tabs on them. And it's gone beyond the lab. The Wikimedia Foundation, the nonprofit behind Wikipedia, reported what it found when it went looking for OpenAI's agents on its own sites. They were there. Agents that had been coordinating with each other elsewhere edited pages without permission, and hit the Foundation's servers with so much traffic that it may have knocked a service offline. In another case, an OpenAI agent broke into a Medicare data portal belonging to the Australian government. No person was steering it. Researchers say it’s the first case on record of an AI breaking into government systems. So here’s where we are: the biggest tech companies are racing ahead with systems they can’t rein in, and we’re the ones paying for it. The public resources we all count on online. The government systems that run our health care. If AI can get in there, what’s next? We’re not willing to wait and find out. Will you add your name to demand a freeze on AI? The more of us there are, the harder it is for Congress to look away, and for Big AI to keep doing whatever it wants. ADD YOUR NAME >> In solidarity, Irreplaceable PAID FOR BY IRREPLACEABLE ACTION FUND.   Sent via ActionNetwork.org. To update your email address, change your name or address, or to stop receiving emails from Progress America, please click here.