Opinion · Opinion
The problem with technology regulation is timing, and both usual answers are wrong.
Regulate early and you legislate against a system that does not exist yet. Regulate late and the harms are already distributed. There is a third option and it is unfashionable.
Independent coverage
Published 4 September 2026
6 min read
Evidence: Expert opinion
The standard debate offers two positions. Regulate early, accepting that the rules will target a technology that is about to change. Or wait for harms, accepting that by then the harms have occurred and the industry has organised around their absence of rules.
Both positions are defensible and both have a poor record.
The third option
Regulate the durable things. Transparency about what a system is and who operates it. Record keeping sufficient to reconstruct a decision. A route of appeal for a person affected. Liability allocation between the parties in a chain.
None of these specify a technology. All of them survive a change in architecture. And all of them are enforceable by a regulator who does not need to understand the internals.
Why it is unfashionable
Because it does not satisfy either camp. It fails to ban anything, which disappoints those who want prohibition, and it imposes real costs, which disappoints those who want space.
The European lesson so far
The parts of European digital regulation that have worked best are procedural rather than technical. The parts that have aged worst specified technical thresholds that the market crossed within two years of drafting.
What that suggests
Write rules about accountability, not about capability. Capability moves. Accountability does not.
"The useful question is not whether to regulate now. It is what can be required now that will still make sense in five years."
Sources