We spent years asking whether artificial intelligence would become powerful.
Now a more practical question is arriving: who gets to decide what powerful AI is allowed to do?
The European Union’s AI Act entered another important phase in August 2026. The European Commission says enforcement powers for certain provisions began on 2 August, including rules covering prohibited AI practices, general-purpose AI obligations and transparency requirements for certain systems.
Among other things, some AI systems must make it clear when a person is interacting with AI, while certain synthetic content must be labelled.
Whatever your view of regulation, there is a bigger issue here.
Technology doesn’t arrive in a vacuum.
It enters schools, workplaces, hospitals, banks, government departments and everyday conversations. Once it is embedded in those systems, reversing a bad decision becomes much harder.
That means AI governance isn’t really just about computers.
It’s about power.
Who can make decisions?
Who can challenge them?
Who is responsible when an automated system causes harm?
Who gets access to the technology?
Who is excluded?
And who writes the rules?
The book How We Can Fix Things: Unlocking Solutions for a Better World argues that emerging technologies need to be considered as part of wider systems of governance, employment, equality and public responsibility.
That is particularly relevant with AI.
Suppose an automated system becomes very good at screening job applicants.
It might save an employer time.
But if its training data reflects historical inequalities, it could also reproduce them at enormous scale.
Or imagine AI being used in public services.
It might make decisions faster.
But speed isn’t the same as fairness.
The temptation is to ask whether AI itself is “good” or “bad”.
That’s probably the wrong question.
A hammer can build a house or break a window. What matters is who uses it, under what rules, with what safeguards and with what accountability.
The EU’s approach is one attempt to answer those questions through a risk-based framework.
It won’t settle the global debate.
Nor should it.
Technology changes too quickly for any single piece of legislation to be the final word.
But the principle is important: society doesn’t have to wait until a technology has caused widespread harm before deciding that rules might be useful.
The deeper challenge is creating governance that is fast enough to keep up without becoming so rigid that innovation becomes impossible.
That’s a difficult balance.
Perhaps the most important AI question isn’t “What can this technology do?”
Perhaps it is “What should society allow it to do — and who gets a say?”
#ArtificialIntelligence, #AI, #AIGovernance, #EUAIAct, #Technology, #EthicalAI, #DigitalPolicy, #Innovation, #FutureOfWork, #HowWeCanFixThings
Want more like this?
Get an occasional email when there’s something new worth reading. No spam, unsubscribe anytime.
By subscribing you agree to receive occasional emails about new releases. Unsubscribe anytime by replying to any email.