Students edition
From January 2027, SURF is building its own AI environment for secondary vocational, higher professional and university education, while legal scholars in Tilburg and Maastricht want to keep generative AI out of the lecture hall.
SURF AI arrives in January 2027: your school will soon choose which model you can use
A shared AI environment for secondary vocational, higher professional and university education, with its own chat app and an institution in control.
SURF announced SURF AI on 24 September. From January 2027, SURF members from secondary vocational, higher professional and university education will get access to it. It was built together with Npuls. SURF calls it a safe and responsible AI infrastructure for education and research. SURF AI has two parts. The first is an AI hub that safely connects applications, users and different language models. This includes open-weight models: models whose makers have made the internal values public. The second part is eduGenAI, a chat application that you use just like a normal chat window. Important detail: your institution decides which language models you get access to, based on public values such as autonomy, transparency, privacy and inclusion. For you, this eventually means a chat environment from your own school or university, as an alternative to tools like ChatGPT. Ask your programme or student council whether your institution is taking part. A hub like this is a middle layer: you type your question into eduGenAI, the hub sends it to a language model and passes the answer back. The model itself doesn't change, but the agreements around it do: which models are connected, who processes the data and what may be done with it. Because institutions choose which models are switched on, one programme may offer something different from another. Open-weight models are attractive because an institution can run them itself, instead of sending questions to an outside company. Watch out for three things: January 2027 is still a long way off, so this solves nothing for your exams this year; a safe environment says nothing about what your programme allows for assignments, that's in your study guide; and it's still the same kind of language model, with the same tendency to confidently say something wrong. The gain is in control and privacy, not in better answers.
Background for subscribersThis edition is about who holds the controls when it comes to AI. SURF is building its own chat environment for education, where your institution decides which language models you're allowed to use. At the same time, other stories show that giving access also brings risks: AI agents bypassing security, fake job applications with harmful code, and deepfakes making scams more convincing. And while Google is making learning with AI easier through Study notebooks, law scholars want to keep AI out of the lecture hall altogether. The common thread: the more access AI gets to your data and accounts, the more important it becomes to check for yourself exactly what's happening.
US bill wants to ban 'superintelligence', with prison terms of up to twenty years
Senator Bernie Sanders and Congressman Greg Casar introduced the 'Ban Artificial Superintelligence Act' on 23 September, a 19-page proposal. It wants to permanently ban AI systems that outperform human thinking in most areas. It also wants to temporarily pause the building of advanced AI systems, until a new US Department of AI has drawn up oversight and safety rules, with its own AI secretary. Breaking the rules would carry the 'corporate death penalty', plus up to twenty years in prison: comparable to penalties for illegally building nuclear weapons. The proposal comes as Chinese President Xi Jinping also told Trump that AI must stay under human control. So AI oversight is now on the agenda between the US and China too. Follow this if you study something related to technology, law or governance: it will shape the playing field of your subject. Banning 'superintelligence' first requires a definition. This bill chooses a performance threshold: systems that outperform humans on most thinking tasks. It opts for a pause beforehand instead of punishment afterwards, with oversight from a new federal department. That differs from the European AI Act, which sets rules per risk level instead of banning a whole category. Note: introducing a bill isn't the same as passing a law. There are only two sponsors, and it only applies in the US. What matters most for you is the direction: oversight of the heaviest models is becoming a major political topic, even in countries that have mostly focused on speed so far.
Background for subscribersThis is a taster.
Subscribers read the whole Students edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Students edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026