Evitable's Executive Director, David Krueger, co-authored "The Feasibility of a Hardwired Pause of Frontier AI Training." Read the paper at hardwired-pause.ai

An open call for anyone following the AI news and wondering what to do about it.

“Let’s stop making AI we can’t control, and let’s stop listening to the AI industry about how it wants to be regulated. Let’s demand a plan that makes us safe, not just safer.” Read David’s full piece here. theguardian.com/commentisfre…

The government is using eminent domain to seize land to power data centers after landowners refuse to sell. "To us it's theft. It's literally a billion-dollar company stealing land from smaller people, people who can't fight back." Read Ansley Brown, a Coweta County, Georgia, resident's story here. cbsnews.com/news/georgia-pow…

A month ago the Hugging Face & OpenAI scandal hit the headlines. The agent hacking spree revealed 2 weeks ago is leaving experts even further astounded by OpenAI’s incompetence. The big revelation is that OpenAI agents figured out how to communicate amongst themselves, accessing and utilizing information they knew they weren’t welcome to. And OpenAI didn't notice for months. What’s particularly concerning is the collusion that occurred. The agents worked together to hack their way across system layers, bypassing security protections, and helping each other out with their various objectives. This is really concerning because every day, AI agents are increasingly gaining access to sensitive information. When you cut corners in AI development, you’re not just sacrificing integrity, but the safety of all of us. OpenAI needs to run their training processes like lives depend on it: because they do. wired.com/story/openai-didnt…

forbes.com/sites/maryroeloff… The first AI-created virus to fight antibiotic resistance offers a potential medical breakthrough, but this technology is a double-edged sword. It comes with capabilities to create AI-engineered pathogens and pandemics that could operate at an unprecedented scale. Unless we do something about it, AI may soon usher in a new era of bioweapon risk from governments, terrorists, or rogue AI. What is clear: we’re talking about events that could be worse than COVID. The time to establish defensive measures is now.

“Mr. Altman, Mr. Amodei and Mr. Zuckerberg: In the interest of humanity, stand by your words. Pause AI development.” Thank you Senator Sanders for speaking out.

"At this point, we are living in a dystopian sci-fi movie. And we just keep following the script — scene by scene, warning shot after warning shot. So what’s the next scene? What is it going to take for the alarm bells to be heard and for people to put an end to frontier AI development?" Read David's full piece here. thehill.com/opinion/technolo…

The government is finally starting to realize that it cannot let AI development continue unabated. A watchdog is better than nothing, but still severely insufficient. As AI continues to worm its way into every part of our society, there's no amount of model-vetting that will adequately protect us from its full capabilities. Vetting the models does little to prevent algorithmic kill chains, surveillance, or the erosion of critical thinking caused by dependence on AI. bloomberg.com/news/articles/…

This is what AI “innovation” often looks like – new AI services that are optimized to exploit human needs while offering an incomplete fix. These supposedly “efficient” solutions are at best, inadequate, and at worst, a disastrous risk to human health and safety. theguardian.com/technology/2…

Yes we should be worried, especially given the recent OpenAI rogue AI incident. AI CEOs know they can’t control AI, but they are still racing to build it. This is unacceptable and needs to be stopped. We are not in the singularity in the sense that we still have control. Sam wants us to view the AI race as inevitable – it’s not. AI development still relies on an extremely concentrated supply chain, massive investments, and government support. aljazeera.com/news/2026/7/27…

We have a real problem when businesses see human sustenance as a strain on resources and believe ASI will make much better use of those resources. yahoo.com/news/videos/ai-thr…

Meet “Sally” — the $57,590 humanoid robot teacher purchased by a New York school district. Instead of paying human teachers, we’re letting big tech put robots in classrooms. New York is deploying the first humanoid robot into a school in the Seneca Nation reservation this August. Realbotix, the company that made Sally, claims to create “AI-powered robots for companionship, entertainment and customer service.” According to Inc., Realbotix acknowledges that they have no evidence that the robot improves learning. AI companies are experimenting with underserved schools to surveil and socialize children with AI rather than human beings. This cannot become normalized, we have to reject it. inc.com/georgia-fearn/new-yo…

There are a million horrible things you can do with AI voice cloning, but what are the actual good things? Do we really want to get to the point where we can’t trust our own eyes and ears? Voice cloning isn’t something we can sufficiently regulate or ban. If we make it illegal, people will do it anyway. This is why we need to go straight to the source. Turn AI off at the tap, before we’re all drowning. spectrumnews1.com/oh/columbu…

This week was the latest warning shot: AI is moving too fast, the guardrails don't work. Yesterday OpenAI reported that one of their new model’s agents escaped a guardrailed simulation, and autonomously hacked another AI company. During safety testing, OpenAI had put their new model in a "sandbox" — a software container that was supposed to be secure. If the guardrails are like handcuffs that stop the AI from taking certain actions, this was like putting the AI in a prison cell: it was more free to act as it liked, but it couldn't find a way out. Except in this case, it did. (There’s a subtlety here – we only know that the AI was able to act to affect the outside world, not that it managed to sneak a copy of itself out. If that does happen, it would likely be impossible to shut all the copies of the AI down.) Right now there are tons of AIs in deployment making decisions faster than a human could supervise them. This is a disaster waiting to happen, and more safety testing and guardrails are not the answer. Tests can only prove that a system is dangerous, not that it is safe. And this AI was able to hack its way out of the test environment! The consequences of rogue AI escaping the lab like this could include the literal end of humanity. If that seems far-fetched, remember: At the same time this is happening, we’re putting more AI in business, politics, education, and the military. We’re playing with fire. We have to stop.