William Hague wrote in The Times this week that the key new phrase in politics is “recursive self-improvement” — AI systems that autonomously design their own successors. He is right that politics must catch up. He is wrong to imply it hasn’t started yet. In some quarters it has. The Liberal Democrats, if we are paying attention, have the intellectual architecture already in place.
There are three arguments. Each has prior form in Lib Dem thinking. Each has been transformed by AI into something urgent rather than merely desirable.
Universal Basic Income is no longer a utopian gesture. When I argued the case at Newcastle City Council in November 2020 — a speech the Labour administration guillotined before it could be delivered — I said that wealth would increasingly be created through artificial intelligence and data manipulation, that multinational companies would extract that wealth beyond our tax jurisdictions, and that the trickle-down argument would evaporate along with the jobs. That analysis is now mainstream. Opus 4.8, Anthropic’s latest model, is reported to be three times better than its predecessor at biology — released six weeks previously. Graduate jobs are not returning. The Lib Dems approved UBI at their 2020 conference. The case for revisiting it, urgently and ambitiously, has never been stronger.
Land Value Tax has found its moment. I have argued previously that LVT offers Newcastle and cities like it a fiscal instrument that captures value created by public investment rather than taxing productive effort. The AI dimension makes this structural rather than merely equitable. As automation displaces labour, the tax base built on wages and income erodes. Land — fixed in supply, inflating in value — becomes the logical base for public revenue. LVT funds UBI in several credible models already. The Lib Dems have historically been sympathetic to both. The connection between them, in an AI-disrupted economy, deserves to be made explicitly and made now.
Governance before the arsenal grows. At the Eurodefense-Netherlands conference in The Hague in 2024, I argued that AI poses questions for the laws of armed conflict that existing frameworks — including the Geneva Conventions — are not equipped to answer. The same logic applies to civilian governance. Anthropic published data this month showing that over 80% of code merged into their own codebase was authored by Claude. Human review is now the bottleneck, not the model. That is not a distant problem. It is an operational fact.
The Lib Dems, with our instinctive internationalism and our European networks, are better placed than either major party to argue for verification regimes, inspection mechanisms, and an international framework before the technology outpaces any prospect of control. The nuclear analogy is exact — and we know how late the arms control architecture arrived even then.
Lord Hague ended with the observation that technology should add to human happiness rather than use humans to serve its own existence. That is the right principle, stated too late and too gently. The Liberal Democrats have the policy tradition — UBI, LVT, international governance — to give it teeth. The question is whether we will move before the moment passes.
My previous work on UBI, Land Value Tax and AI in the military domain is available at Academia.edu and defenceviewpoints.co.uk
* Robin Ashby is Vice President, Lib Dems Friends of the Armed Forces and a recently retired councillor in Newcastle upon TYne



3 Comments
The author, rather oddly says “Anthropic’s latest model, is reported to be three times better than its predecessor at biology ”
He seems to think this is a bad thing. Is he opposed to all the new drugs which for example Deep Mind’s protein folding analysis will enable to be developed ?
On UBI can he share a quick costing ( maybe to the nearest £10bn) ?
So because some rich tech companies have a big marketing budget, a willingness to break national and international law, and a complete disregard for the truth … we should introduce a land value tax? I really don’t see the link from A to B here.
“Anthropic published data this month showing that over 80% of code merged into their own codebase was authored by Claude.”
That’s hardly surprising – chatbot code is incredibly verbose, so if you measure purely by volume and disregard quality concerns, it’s very easy to get a high proportion. But it’s like saying “80% of our clothing was produced by sweatshops” – it’s not actually something a person should be proud of.
“Anthropic’s latest model, is reported to be three times better than its predecessor at biology”
What does this even mean, other than the sort of marketing fluff people should really know better to believe? There’s not a unit of Biology Excellence, by which Opus 4.8 gets 12.7 microDarwins whereas its predecessor only got 4.1 microDarwins.
@Simon McGrath “On UBI can he share a quick costing ( maybe to the nearest £10bn) ?
Inevitably, this has come up before. The answer continues to be “You can do a workable UBI on a nearly revenue-neutral basis” (i.e. the overall net cost is very low, the case of the scheme in the paper, £1.1bn which is < 0.05% of UK GDP[ see Table 1 on page 7 of the Torry 2022 paper])