Something I am acutely aware of while using Claude Code is how actionable the body of all human knowledge has become. The internet made it accessible. AI made it operable.
AI didn’t come from nowhere. The large language models underlying today’s tools are trained on centuries of human writing and art produced by millions of people. Their real world utility is built on open source coding languages and software that was never intended to be used in this way. People of all types will claim their work is being stolen, and it is.
But human progress has always been built on the work of our forebears. Artists have their influences, writers contribute to the canon, researchers start from a literature review, and computer scientists build on old solutions to make new ones. AI doesn’t change that, it just accelerates it at a pace that is uncomfortable.
AI is a privatisation of a commons built over generations. If we do nothing, the gains from exploiting that commons are going to flow at extraordinary scale to those who build and profit from AI. If AI delivers on its productivity potential, it will be the most significant economic step-change in generations. That represents an enormous pool of gains. My question is whether some portion of those gains is captured for the public, for current and future generations, or whether they compound entirely in private hands.
To the winner goes the spoils?
We’ve spent a decade hearing that data is the new oil. When oil was discovered off the Norwegian coast, it was treated as a national commons rather than simply a windfall for the companies extracting it. The returns were invested in a sovereign wealth fund, now exceeding $2 trillion, managed for future generations. When private companies extracted value from a shared resource, Norway chose to capture some of that value for everyone, not just for those who got there first.
The window for that kind of thinking about AI is now. Once gains compound into private wealth — capitalised into assets, passed through estates, locked into structures designed to preserve them — the options narrow considerably, and the vested interests become entrenched. Future generations will inherit a more productive economy, but one where the gains from that productivity are already priced into assets they can’t afford to own.
The question here is structural: the tax system was built for an economy where creating value meant employing people. That assumption is breaking. The economy should serve society. Not the other way around. Firing people makes financial sense: fewer people, lower costs, wider margins, higher share price. But not only does employment generate tax revenue for governments, it provides structure, identity, social connection. And the costs when it disappears don’t land on the company’s income statement. They land on the public’s.
Employment generates tax income that compounds across the economy: personal income tax, payroll taxes, healthcare levies. All flowing to government every pay cycle, together with the taxes on the economic activity those people put their salary towards. Add it up and companies with large workforces are the tax collection engines. Companies without them are not.
What now?
How can we shift the tax system to more equally tax economic value created, easing the tax burden for companies that employ people, and better tuning the incentives to ensure that the economy works for people and not the other way around?
Across all economies globally we need wholesale structural changes that fundamentally overhaul systems designed for the first half of the 20th century. High level, we desperately need to reduce taxes on labour and increase taxes on capital. We need to get rid of out-of-date taxes that disincentivise employing people. Payroll taxes have not been a good idea for a long time.
And we need to do this in a way that doesn’t simply encourage companies to go find a tax haven to operate out of. Future AI driven companies will be highly mobile, employ very few people, and be easily moved to lower taxing jurisdictions. If we’re not careful, the Cayman Islands will become the next Silicon Valley, full of 10 person unicorns.
It will require an international approach that seems completely unrealistic in the world we’re in today. But we have seen in the past that when emergencies happen, things can move fast to break down bureaucratic barriers. I can only hope that international diplomacy can respond with the appropriate collaborative vigour.
The gains from AI are real and built on a foundation that belongs to all of us — centuries of accumulated human effort that nobody was compensated for contributing. Whether some portion of those gains flows to the public, for current and future generations, or whether they flow entirely to private wealth, depends on decisions being made now. The longer we wait, the harder it will be to correct course.
