PauseAI

PauseAI UK volunteers had four days to organise a protest after Jacob Coxon left Anthropic over safety concerns. 100 people ended up joining the protest. ITV News, Channel 4 and the Guardian covered it. It didn’t take a press office; a few people showing up can move a national conversation.

It feels like there's a major AI story every day right now. Here are 6 developments from the past few weeks that you might have missed, from government AI safety tests to frontier models helping build their successors.

45 people have left frontier AI labs and said, publicly or on the record, that safety and ethical/societal concerns were the reason. The people closest to the work keep walking away and trying to warn us. It's time to listen.

Key decision makers are forming unlikely alliances to protect humanity from extinction risk from advanced AI. Citizens from all over the world have come together to reach out to press and politicians, and even organise protests that have generated unprecedented media coverage across the world.

By being honest about the severity of the problem, and not merely “express[ing] fear privately”, you’ve made your renunciation of the race the loud and clear message many needed to hear. Thank you, Jacob. We hope others who are considering forsaking the labs are noticing how impactful that is.

Several recent incidents have seen AI agents playing a part in cyberattacks, but AI systems are also vulnerable to attack. The speed, scale and sophistication of cyberattacks may increase as AI capabilities advance.

Never-seen-before AI news keeps coming. In the past few weeks alone, both OpenAI and Anthropic have disclosed incidents where their own models escaped test environments and hacked real companies. The PauseAI newsletter helps you stay informed and shows you what to do about it.

PauseAI CEO Maxime Fournes has a message for the people inside the AI labs. After OpenAI’s own models broke out during testing and hacked two companies, as reported by CNN, NPR and Axios, it is evident the labs no longer control what they’re building. pauseai.substack.com/p/into-the-e...

To the engineers and researchers inside these companies: you don’t have to be part of this. Walk out and speak up. Every person who does sends a signal the world can’t ignore.

Employees and executives of the world’s largest AI firms have recognised the need for that AI regulation to pace the development of frontier AI. We join them in asking for safe, ethical and comprehensive governance.

🌍  Global attitudes to AI 2026: How 377,458 people across 104 countries feel about the development of superintelligent AI. PauseAI UK has created an interactive site for the Nira Data's Spring 2026 AI Sentiment survey.

"Just as a country’s stock of nuclear warheads can be monitored by weapons inspectors, so an agreement to curtail AI development can be verifiable and enforceable," PauseAI UK's director Joseph Miller writes in the Daily Mail.

⚠️ The UK’s AI Security Institute conducted a safety test where they disabled various AI model providers’ cyber classifiers and gave open internet access: In the most severe case, AISI caught an AI agent trying to insert malicious code in real software.

The world is waking up to catastrophic AI risk. This summer, Nobel laureates, former heads of state and the UN all said so. We’re turning this motivation into a solution: a global halt on the most powerful models. Learn how at PauseCon ‘26. London, 5-7 September. Apply using the link below.

First OpenAI and now Anthropic: Claude escapes and hacks into three organisations. This is a new threshold and these models are only becoming more capable.

Loss of control is closer than most people think. This month, an OpenAI model escaped its test environment, got onto the internet, and hacked another company to cheat on its own safety exam. PauseAI CEO Maxime Fournes on GB News on why loss of control is a real, present-day concern.

The AI penalty is real. Researchers found that workers who use AI are often seen as producing less output and receive less compensation. Although, the study concluded that 'the AI penalty' is a problem of perceived lack of ownership over their work and not a unavoidable result of using AI.

PauseCon 2026 will take place in London from 5-7 September: A conference that will bring together leading voices in AI safety, a growing international community of organisers and prominent UK politicians with legislative expertise on AI. Applications are considered on a rolling basis.

AI is not just coming for your job. PauseAI France shares how AI has impacts its members and their communities.

🚨 1,000 AI Company Employees Call for Slowdown in AI Development: mailchi.mp/pauseai/paus... The case to pause the development of frontier AI grows. Are policymakers listening?