Stuart Russell, one of the world’s leading AI researchers, has weighed in on the growing calls from AI lab leaders to “pace” frontier AI development. His starting point is Dario Amodei’s recent proposal that the major AI companies should slow the rate of progress enough to give safety research more time to catch up. Amodei has proposed independent evaluators working inside AI labs, common safety standards across frontier companies, and eventually international coordination. Russell agrees that the industry has a serious problem, but he thinks the “pacing” idea gets the order wrong. His analogy is aviation. We would never say: “Boeing can release a new aircraft every year, and hopefully there is enough time to complete the safety testing before each one arrives.” Instead, an aircraft is allowed to fly only once it has passed the required tests and been certified as safe. Russell argues AI should work the same way. Rather than deciding how quickly models are allowed to improve, regulators should set clear safety requirements first. If a model reaches a certain capability level, developers should have to prove that it also meets specific alignment and safety standards before it can be deployed. And if they cannot prove that? Progress stops. Russell calls this the “red lines” approach. In his Formula 1 analogy, what AI may eventually need is not a safety car slowing everyone down, but a red flag stopping the race until the conditions are safe. His concern is particularly focused on recursive self-improvement — AI systems becoming increasingly capable of improving themselves — and the possibility of eventually creating systems more intelligent than humans without a credible way of ensuring humans remain in control. Russell’s argument is therefore stronger than “we should be more careful”. It is: Capability should never be allowed to outrun demonstrated safety. And if the safety problem cannot be solved, he believes there may come a point where further development simply has to stop, regardless of how much money has already been invested.

The biggest news from recent weeks is around the need to pace AI development and build an environment where the work of all AI labs can be monitored. In his article in The Guardian (a UK sunday newspaper) he wrote a piece headed "AI safety requires more than just slowing our pace" In it he argues that slowing down must be preceeded first by making sure it is safe. I cover this in my article "Stuart Russell: slowing AI down isn’t enough — safety has to come first" elsewhere in the community. I cover the news about safety and slowing down in this week's AI News in the classroom now.

If you've been following the news recently, you may have noticed a definite change in the conversation around Artificial Intelligence. For the past few years, most of the excitement has been about what AI can do. Better ChatGPT models. AI-generated images and video. Coding assistants. AI agents. Medical discoveries. Robots. Self-driving vehicles. Tools capable of doing in seconds what might previously have taken us hours. I've certainly been excited about many of those developments myself. But during September 2026, something rather interesting has happened. Some of the people at the very heart of developing the world's most powerful AI systems have started saying, very publicly: We may be moving too quickly. And I think that's something worth taking seriously. Not panicking about. Not assuming that some science-fiction-style robot takeover is just around the corner. But certainly taking seriously. What's Changed? We've heard warnings about the potential dangers of AI before. People have been talking about Artificial General Intelligence, superintelligence and even the possibility of AI eventually becoming an existential threat to humanity for years. It has therefore been quite easy to dismiss some of it as speculation. What makes the current discussion different is that the warnings are increasingly coming from researchers and executives actually developing frontier AI systems. On 12th September, Dario Amodei, the CEO of Anthropic — the company behind Claude — published an essay called We Must Pace the Frontier. Amodei is by no means anti-AI. Quite the opposite. He believes advanced AI could help cure major diseases, dramatically increase economic growth and improve human life in ways we can currently only imagine. But he now believes the pace of development itself has become a problem. He argues that the industry needs to give AI safety research more time to keep up with improvements in AI capability. Among the risks he identifies are losing control of advanced systems, cyberattacks, biological misuse and serious economic disruption.

A recent news item in the Free Press Journal raises an interesting point about the control of AI development that many experts are now calling for. They report that OpenAI CEO Sam Altman told employees that the company may slow the pace of advanced AI development and seek coordination with other labs. The move comes after growing safety concerns among AI researchers, including warnings from OpenAI scientists and former employees of leading firms. However, any industry-wide agreement could face antitrust challenges. OpenAI is considering slowing down the pace of its cutting-edge artificial intelligence (AI) development, and chief executive Sam Altman is hoping other AI companies will follow suit. What Altman told staff? In a companywide meeting this week, Altman told employees that OpenAI could potentially pace its AI development, possibly in coordination with several other AI labs, though he acknowledged that some companies may not agree to do so, Bloomberg reported. The company has already slowed parts of its model development and paused certain internal AI training in recent months over safety concerns. The reported shift comes amid a surge of public unease from within the industry's own ranks. In recent weeks, employees at leading AI firms have raised concerns about the escalating risks of advanced AI systems, fuelling anxiety inside these companies, Bloomberg reported. OpenAI's chief scientist, Jakub Pachocki, recently wrote a post warning about the dangers of AI, saying he believes AI companies should be coordinating to slow down future development. The unease boiled into public view this week when Jacob Coxon, an AI researcher, quit his job and accused both of his former employers, Anthropic and OpenAI, of 'gambling with our lives' by racing toward superintelligent AI. Coxon said in a social media post that people building AI systems believe the technology could "kill us all by the end of the decade," a comment that went viral, drawing more than 150 million views along with attention from lawmakers and public figures.