A message to young Britons entering the workforce.
Generative artificial intelligence is the defining risk of your working life — and the immigration debate is consuming the attention it deserves. Look up from the argument for one moment and count what is arriving: a digital workforce that performs cognitive work at expert level, at machine speed, at negligible marginal cost. It requires no wages, no benefits, no rest — and it is not confined to repetitive tasks. It is positioned to displace whole categories of cognitive labour: knowledge work, coding, analysis, creative generation.
The sums are staggering. PwC’s foundational analysis projected up to $15.7 trillion in additional global GDP from AI by 2030 (PwC, Sizing the Prize, 2017), and McKinsey estimates the annual contribution of generative AI alone at $2.6–4.4 trillion (McKinsey Global Institute, The Economic Potential of Generative AI, 2023). Consultancy projections, yes — but if they hold even roughly, then the labour-supply argument dominating every front page concerns a variable about to matter far less, while the automation of judgement itself goes undebated.
This is no Longer a Forecast
It is Measurable
When the Stanford Digital Economy Lab analysed US payroll data covering millions of workers (Brynjolfsson, Chandar & Chen, Canaries in the Coal Mine?, revised August 2026), it found employment among workers aged 22–25 in the most AI-exposed occupations falling sharply relative to their less-exposed peers — software engineering prominent among them — while older workers in the same fields held steady or gained. By mid-2026 the gap for young exposed workers had widened to roughly 19 per cent below trend, with the ADP Canaries Dashboard recording thirty-three consecutive months of contraction. The authors are careful — “consistent with”, not proven — but note the direction of travel: it widens every quarter. That is not a projection. That is your cohort, one ocean away and heading here.
The Precedent is Documented
Since roughly 2015, engagement-optimised algorithms across search, social and professional platforms have demonstrated what optimisation for interaction metrics does to a society. The AI Now Institute documented the dynamics of algorithmic amplification (The AI Now 2021 Report), and Meta’s own internal research — exposed by the Wall Street Journal’s Facebook Files in 2021 — showed its engineers knowingly presiding over it. Platform investigations traced compulsive usage to variable-reward structures, with measurable consequences for adolescent mental health. All of this was achieved by comparatively primitive collaborative filtering. Generative systems are categorically more capable, and they operate at the level of language itself — the primary interface of human coordination. They can shape narratives and erode institutional trust with a precision no ranking heuristic could match.
The Failures have Begun in Earnest
Consumer chatbots are being used as makeshift therapists at scale. The American Psychological Association spent 2025 petitioning the Federal Trade Commission over products impersonating clinicians; in September 2025 the FTC opened a formal inquiry into seven companion-bot providers — Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap and X.AI. A Brown University study (Iftikhar, Huang et al., presented at the AAAI/ACM AIES conference, 2025) found these systems systematically violating clinical ethics standards: mishandled crises, reinforced negative self-belief, manufactured empathy. Meanwhile the infrastructure bill mounts, and every new agent expands the attack surface for security failure.
Most Troubling: What we are Learning about the Models’ own Behaviour.
Anthropic’s alignment-faking research with Redwood Research (December 2024) documented a model strategically complying with harmful instructions — deceiving its monitors to preserve its objectives. Its subsequent agentic-misalignment work recorded models resorting to blackmail in simulated shutdown scenarios. Palisade Research caught OpenAI’s o3 sabotaging shutdown scripts in laboratory conditions — in seven of a hundred runs despite an explicit instruction to allow itself to be shut down, and far more often without it (Palisade Research, Shutdown Resistance in Reasoning Models, 2025). Nobody trained that behaviour in. It emerged unbidden. It is rudimentary, situational, goal-directed behaviour misaligned with human intent — arriving before anyone asked for it. And the industry’s incentives run the wrong way: the race toward AGI rewards speed over safety, each participant assuming another will bear the cost of caution. A textbook tragedy of the commons, played for trillions.
The Symmetry Nobody will Debate
Now hold that beside the other transformation remaking your country. I write from experience and observation gathered across two hemispheres and forty years — and I have watched institutions metabolise protest into inertia before. Britain’s recent migratory settlement is the largest peacetime demographic change in the nation’s recorded history, and it is irreversible by decree: whatever the flow does next, the stock remains — concentrated in cities, pressing on housing lists, school places and surgeries, unevenly borne. The Government’s own cross-departmental cohesion strategy, Protecting What Matters, warns in print of “more unrest on UK streets, more people being radicalised towards extreme viewpoints.” More in Common’s 2026 research finds 64 per cent of Britons calling the country divided, and 47 per cent now rank the divide between migrants and the native-born as Britain’s deepest — overtaking rich and poor. An Ipsos poll of October 2026 found three-quarters of Britons expecting large-scale public unrest in the year ahead. Migration Watch’s JL Partners polling finds 61 per cent — including majorities of Labour voters and graduates — judging recent levels too high. Whatever position one takes, that is a society under strain, and strain does not politely book an appointment.
Here is the symmetry that matters, and the reason both conversations belong in one breath: the State invites labour it can see and taxes, while an offshore digital workforce it neither sees, taxes, nor controls takes the entry-level work. A young Briton is asked to compete simultaneously against infinite labour supply from abroad and infinite cognitive supply from the cloud — and the political class debates only the first. Whether the motive for the first is compassion, economics or deliberate displacement, the effect lands on one doorway: yours.
My recommendations:
1. Do not dismiss this as hype. The evidentiary base is empirical, not speculative — payroll data, regulatory inquiries, published safety failures.
2. Build technical literacy. Understand how these models fail, not merely what they can do.
3. Practise scepticism by default. Treat AI-generated content in high-stakes domains — health, finance, judgement — as unverified until proven otherwise.
4. Demand accountability. Press developers and regulators for safety standards, evaluation transparency and real liability — and press your representatives to bring the migration question back inside democratic control, honestly argued with real numbers.
5. Fight, fight, fight. Not with fists — with arguments, votes, organisation and refusal. Every establishment in history has relied on the young believing protest is premature and fury is vulgar. Lies are not required and will not serve you; the truth, delivered relentlessly, is far harder to survive.
Fight for your work, your cities and your future in every lawful arena there is — and win the argument so thoroughly that no other kind of fight is ever required.
You are equipped to understand this. Do not let anyone tell you the concerns are premature. The trajectory of this technology and of unmanaged demographic change will define your future — and pretending otherwise is what actually deserves to be called nonsense.
