The phrase “AI should remain under human control” sounds like something from the opening ten minutes of a sci-fi movie.
In 2026, it has become a real international policy discussion.
A group of 20 countries, alongside the European Union, has backed a call for stronger global cooperation around advanced artificial intelligence, including the idea that increasingly capable AI systems should remain subject to meaningful human oversight.
Supporters include countries such as Germany, Canada, Australia, South Africa, Singapore and the United Arab Emirates.
The group is also calling for discussions around an international institution that could help establish common standards for advanced AI.
It’s easy to hear “international AI standards” and immediately mentally log out.
But the timing does matters.
For most people, the first big wave of generative AI meant opening ChatGPT and asking it to write something, generate an image or explain homework.
The next generation is different.
Companies are developing AI agents capable of navigating websites, using software and carrying out multi-step tasks with less supervision. Meta’s Muse, for example, is designed to act on users’ behalf online. Other companies are racing toward similar systems.
That makes the question of control much more practical.
If an AI recommends the wrong restaurant, that’s annoying.
If increasingly autonomous systems begin handling financial transactions, accessing sensitive services, operating infrastructure or making consequential decisions, “who is responsible?” becomes a much bigger question.
Countries still disagree significantly about what AI regulation should look like, and an international declaration doesn’t automatically create enforceable global rules.
But the direction of the conversation is revealing.
For years, the tech industry’s favourite question was: How powerful can we make AI?
Governments are increasingly asking a different one:
How powerful should we allow it to become without a human holding the steering wheel?
