Former UK Prime Minister Sunak Warns AI Is Already Cutting Job Opportunities for Young Workers

Rishi Sunak, the former UK Prime Minister who staked part of his political legacy on AI governance, told the BBC in early 2026 that artificial intelligence is "flattening the jobs market for young people" [1]. CEOs, he said, are privately telling him that "flat is the new up" — they can grow revenues without growing headcount [1]. His proposed remedy: eliminate National Insurance — a payroll tax — to make hiring humans more competitive against software [1].

The warning landed in the middle of mounting statistical evidence that young workers are bearing the earliest and steepest costs of AI adoption. But it also came from a figure whose ties to the AI industry, and whose political incentives, complicate the message. This investigation examines what the data actually shows, who stands to lose, who disagrees, and what — if anything — governments are doing about it.

The Numbers: Entry-Level Hiring Is Falling Fast

US entry-level job postings have declined roughly 35% since January 2023, according to an analysis by Rezi.ai drawing on job board and employer data [2]. The drop is far steeper in specific sectors: junior tech and software development postings fell 67%, and in the UK, computer programming roles for 16-to-24-year-olds collapsed by 44% in a single year [3]. Across major EU countries, LinkedIn and Indeed data showed a 35% decline in junior tech positions through 2024 [2].

Source: Rezi.ai / Dallas Fed / LinkedIn Data

Data as of Mar 1, 2026CSV

A January 2026 study by the Federal Reserve Bank of Dallas provided one of the strongest empirical signals yet. In occupations with the highest exposure to AI — defined by the share of tasks that large language models can perform — employment for 22-to-25-year-olds fell 6% between late 2022 and July 2025 [3]. Over the same period, employment among workers aged 30 and older in those same occupations grew 6 to 13% [3]. The pattern suggests that AI is not simply eliminating jobs but redirecting them away from the least experienced workers.

A survey of enterprises found that 66% are actively reducing entry-level hiring because of AI capabilities [2]. The World Economic Forum's 2025 Future of Jobs Report estimated that 40% of employers expect to reduce staff where AI can automate tasks [4].

Who Is Cutting, and How Many Jobs?

More than 50,000 layoffs in 2025 were explicitly attributed to AI by the companies involved, according to CNBC's tracking [5]. That figure sits within a broader wave of 127,000 tech-sector mass layoffs during the year [5]. Fortune reported that 66% of CEOs acknowledged freezing hiring while directing investment toward AI, and that corporate America eliminated more than 1.17 million positions under the logic that labor costs had to be cut to fund AI deployment [6].

Source: CNBC / Fortune

Data as of Jan 1, 2026CSV

Specific disclosures include Amazon (14,000 corporate roles, with CEO Andy Jassy stating "some jobs will no longer be needed because of AI tools"), Microsoft (approximately 15,000 roles), Intel (24,000 as part of a restructuring to 75,000 employees), Dell (12,500 as it reorganized around AI server demand), and Salesforce (over 4,000 roles, citing AI's capacity to manage customer service and development tasks) [5][6].

Harvard Business Review raised a critical nuance: many of these layoffs are based on AI's potential, not demonstrated performance [7]. Companies are cutting headcount in anticipation of productivity gains that have not yet materialized, creating a gap between the AI capabilities that executives project and the AI capabilities that currently exist.

Youth Unemployment: The Headline Numbers

UK youth unemployment has reached 16%, with 732,000 people aged 16-24 out of work — nearly 100,000 more than a year ago and the highest rate in a decade [8]. For the first time, the UK's youth unemployment rate has overtaken the EU average of 14.7% [8]. The British Chambers of Commerce forecasts the UK figure rising to 17% in 2026 [8].

Source: ONS / BLS / Eurostat

Data as of Apr 1, 2026CSV

In the United States, youth unemployment stands at 10.8% for 16-to-24-year-olds, compared with 4.3% for the general workforce [9]. The overall US unemployment rate has edged up from 3.4% in April 2023 to 4.3% as of March 2026 [9].

Source: FRED / Bureau of Labor Statistics

Data as of Mar 1, 2026CSV

Source: ILOSTAT — International Labour Organization

Data as of Dec 31, 2024CSV

A King's Trust survey found that 74% of 16-to-25-year-olds in the UK worry there will not be enough jobs for their generation, while 59% are specifically concerned about AI's impact on their employment prospects [10].

Disentangling AI From Everything Else

Attributing youth employment trends to a single cause is analytically hazardous. The period since 2022 has also seen aggressive interest rate rises by the Bank of England, the Federal Reserve, and the European Central Bank, which suppressed hiring across rate-sensitive sectors including tech, real estate, and finance. Post-pandemic labor market normalization has reversed the unusual hiring surges of 2021-22. And sector-specific downturns — the advertising recession, media industry consolidation, venture capital pullback — have independently reduced entry-level openings.

The Bank of England's Bank Underground blog, in a January 2026 analysis titled "Generative AI: degenerative for jobs?", examined whether observed employment shifts could be attributed to AI or to these confounding factors [11]. The Dallas Fed study addressed this by comparing workers within the same occupations across age groups — finding that the divergence between younger and older workers appeared specifically in AI-exposed roles, not across the board [3]. That pattern is harder to explain through interest rates or post-pandemic correction alone, since those forces would affect all age groups within a given occupation more uniformly.

The LSE's British Politics and Policy blog published an academic assessment asking directly "Is AI behind youth unemployment?", noting that while correlation is clear, establishing firm causation remains difficult given the simultaneous macroeconomic pressures [12]. Economists broadly agree that AI is a factor, but dispute whether it is the primary one or a contributor among several.

The Optimist Case: Why Sunak May Be Premature

Goldman Sachs research points out that 60% of workers today hold occupations that did not exist in 1940, implying that 85% of employment growth over the past eight decades has been driven by technology-created new categories of work [13]. The historical pattern — mechanization, electrification, the personal computer, the internet — has consistently produced net job creation after an adjustment period, even when specific occupations were destroyed.

Morgan Stanley examined five major innovation waves from the Industrial Revolution to the internet era and found that, after two years, none had produced a noticeable impact on total employment [14]. Each displaced workers in specific roles while creating demand for new ones. The bank's analysts argue that AI's labor market impact is likely to follow this pattern.

The Centre for Economic Policy Research offered an additional counterpoint: AI may actually reduce earnings inequality because it disrupts higher-paid white-collar jobs more than low-wage manual work [15]. If a paralegal's tasks are automated but a plumber's are not, the college wage premium shrinks. Early evidence suggests that the lowest-skilled workers derive proportionally greater productivity gains from AI tools than their more experienced counterparts, because the tools raise the floor of what a novice can produce [15].

The Yale Budget Lab's review of the evidence cautioned that researchers "have been terrible at predicting the effects of new technologies on work" and that even AI researchers acknowledge they are making "useful but fallible best guesses" about displacement timelines [16].

However, these optimistic precedents carry a critical caveat. The transition periods during past technology waves lasted years to decades. Workers displaced by mechanization in the 1950s did not benefit from the service economy jobs of the 1970s. Young workers entering the labor market now cannot wait a decade for new job categories to materialize. The question is not whether new jobs will eventually appear, but what happens to the cohort that falls into the gap.

Who Gets Hurt Most

The impact is not evenly distributed. Workers without university degrees face a dual disadvantage: they are more likely to hold roles exposed to automation and less likely to have the educational background to transition into AI-adjacent work [17]. Brookings Institution analysis found that AI increases income gaps through "disparity of skills and access to technology" [17].

Geographic concentration compounds the problem. Entry-level knowledge-work jobs cluster in major cities — London, New York, San Francisco, Berlin — where the cost of extended unemployment or retraining is highest. Workers in regions with less diversified economies face fewer fallback options.

The PNAS Nexus analysis of generative AI's socioeconomic effects noted a paradox: AI offers tools for personalized learning and upskilling that could theoretically benefit disadvantaged workers, but in practice it "may widen the digital divide" because access to training, hardware, and institutional support is unevenly distributed by class and geography [18].

McKinsey's UK analysis described AI's employment effects as "not yet productive, already disruptive" — meaning that the job losses are arriving before the productivity gains and new job creation that AI proponents promise [19]. The brunt falls on workers least equipped to absorb the shock.

The Policy Response So Far

Governments have begun to respond, though the scale of intervention remains modest relative to the problem's scope. The UK government announced a £750 million (approximately $965 million) plan to create apprenticeships and place 50,000 young people into roles in AI, hospitality, and engineering [20]. The program targets the most immediate gap, but 50,000 placements is small against 732,000 unemployed young workers.

In the United States, Executive Orders 14277 and 14278 directed federal agencies to advance AI education and skilled trades training [21]. The broader America's AI Action Plan committed $32 billion over five years for workforce development, including an AI Workforce Research Hub at the Department of Labor [22]. The Senate's AI Workforce PREPARE Act (S.3339) proposed additional funding for retraining programs [23].

Georgetown University's Center for Security and Emerging Technology evaluated the implementation of these executive orders and found early-stage progress but significant gaps between authorization and actual delivery of training programs [24]. Brookings published a critical assessment of worker retraining as a policy response, arguing that historical retraining programs have a weak track record and that the pace of AI adoption may outrun the pace at which workers can be reskilled [25].

Sunak himself proposed eliminating National Insurance — the UK's employer-side payroll tax — to reduce the cost of hiring humans relative to deploying AI [1]. The logic: if a software tool costs less than an employee's tax burden, employers have a financial incentive to automate even when the human worker would produce better results. Removing that tax wedge, Sunak argued, would tilt the equation back toward hiring. Critics note that National Insurance funds the NHS and state pensions, making its elimination a fiscal challenge with consequences far beyond labor markets.

Sunak's Position: Credibility and Conflicts

Sunak's warnings carry weight partly because of his proximity to the AI industry. As Prime Minister, he convened the Bletchley Park AI Safety Summit in November 2023 and established the UK AI Safety Institute — the first government body dedicated to testing frontier AI models before deployment [26]. He has since positioned himself as an advisor to corporate leaders on AI strategy, telling a Goldman Sachs audience that "some jobs will go, but many more will be redesigned" [27].

That proximity also invites scrutiny. The AI Now Institute's executive director warned of "complete industry capture of this conversation," arguing that allowing tech executives to shape regulatory frameworks would "further entrench industry interests" [28]. Civil society organizations were excluded from the Bletchley Park summit, gathering instead at a separate venue [28]. Sunak's post-political career advising corporate CEOs on AI adoption creates an incentive structure where warning about AI's disruption — while counseling companies to adopt it faster — serves his commercial positioning.

His proposed solution, eliminating payroll taxes, would primarily benefit employers and does not create structural labor protections or direct support for displaced workers. It does not fund retraining, create public employment pathways, or regulate the pace of automation. The K-shaped economy framing he used at Fortune's conference — warning that AI's benefits will split sharply between winners and losers — is accurate as description but thin as prescription [29].

What Comes Next: The Projections

Forrester projects that 10.4 million US jobs — 6.1% of the workforce — will be lost to AI and automation by 2030 [30]. PricewaterhouseCoopers estimates that up to 30% of jobs could be automatable by the mid-2030s [30]. The WEF's Future of Jobs Report projects a net creation of 78 million new roles by 2030, but acknowledges this requires massive upskilling — 59% of workers will need reskilling by that date, and only 77% of employers currently claim to be committed to providing it [4].

The BLS has begun formally incorporating AI's projected effects into its employment projections, a first for the agency and a signal that federal statisticians now consider AI a structural force rather than a cyclical one [31].

The central tension in these projections is timing. The job losses are measurable now. The new job categories that AI proponents expect to emerge remain largely theoretical. Goldman Sachs estimates that AI could displace 6-7% of the US workforce if widely adopted, but describes the impact as "likely to be transitory" [13] — a descriptor that offers little comfort to a 23-year-old who cannot get a first job in 2026.

The Bottom Line

The data supports a qualified version of Sunak's claim. Entry-level hiring in knowledge-work sectors has fallen sharply since large language models became commercially available, and the decline is concentrated among the youngest workers in the most AI-exposed occupations. Whether AI is the primary cause or one factor among several remains genuinely contested among economists, but the age-stratified patterns documented by the Dallas Fed are difficult to explain without AI as a significant variable.

The historical precedent that technology creates more jobs than it destroys is well-documented, but that precedent operates on timescales of years to decades — too slow to help the current cohort of young workers. Policy responses exist but are modest relative to the scale of displacement, and the most prominent proposals, including Sunak's, tilt toward making employers' costs lower rather than building structural support for workers in transition.

What is clear is that the traditional career ladder — where graduates enter at the bottom of a firm and learn on the job — is being restructured. The entry-level role that once served as both a learning ground and a talent pipeline is the exact category of work that AI performs most cost-effectively. The question facing policymakers is no longer whether this is happening, but how fast, and what replaces the first rung for the generation now reaching for it.