The number that makes the headlines
On 1 October Challenger, Gray & Christmas, an American outplacement firm that has been counting companies' layoff announcements since 1989, published its September report. The figure most widely picked up is this one: between January and September companies cited artificial intelligence as the reason for 120,136 job cuts, about 21% of the total. It is the leading reason of the year, ahead of market conditions and closures.1 In 2025 the cuts attributed to AI had been 54,836 for the whole year.2
Read that way, it looks like the start of what many people have feared for three years. The same report, though, contains other numbers that get quoted far less. In September the cuts put down to AI were 3,961, 9% of the month, and AI dropped to fifth place among the reasons. In August they were 3,462, the lowest monthly figure since December 2025.3 Total announced cuts in the first nine months, 573,195, are 39% lower than in the same period of 2025, when layoffs in the federal government weighed heavily. Even leaving out the public sector, the fall is 15%.1
What Challenger counts
To read those numbers you need to know where they come from. Challenger collects companies' public announcements and the notices that US law requires before mass layoffs. It records how many jobs a company says it intends to cut and the reason it gives. If a cut is announced as a range or spread over several years, the announced figure is what counts.1 These are intentions and explanations offered to the public. It is not a count of jobs lost, and the reason is whatever the company chooses to state.
The difference in scale from official statistics is huge. In August, according to the Bureau of Labor Statistics, 1.6 million people in the US were laid off or discharged.4 In the same month Challenger counted 52,881 announced cuts, of which 3,462 were attributed to AI.3 The two series measure different things, but the comparison gives a sense of proportion: the cuts companies explain with AI are a tiny fraction of the normal turnover of the American labour market.
How they are spread over time matters too. A third of this year's AI-attributed cuts came in a single month, May, with 38,579 announcements, mostly in technology.2 Technology on its own accounts for 29% of all cuts in 2026.1
Job cuts announced in the US citing AI, month by month
A convenient label
In January Oxford Economics put in writing a suspicion that had been going round for months. In the first eleven months of 2025 the cuts attributed to AI were about 55,000, 4.5% of the total, and in the economists' view even that share was probably inflated: saying you are cutting because of AI "conveys a more positive message to investors" than admitting to weaker demand or to having hired too many people. If AI were already replacing labour at scale, they added, productivity would be speeding up, and it is not.5
In April Andy Challenger himself, commenting on cuts in technology, said the sentence that to me sums up the issue best: "Regardless of whether individual jobs are being replaced by AI, the money for those roles is."6 That is often how it works. The company moves budget from staff to AI investment, and AI becomes the reason even when no program is doing the work of the people who leave.
The big companies show how elastic the label is. At the end of October 2025 Amazon announced 14,000 cuts, presenting them as an adjustment to the new technology; two days later its CEO Andy Jassy told analysts they were "not even really AI-driven, not right now at least", but a matter of company culture.7 Marc Benioff, head of Salesforce, said in September 2025 that he had gone from 9,000 to about 5,000 people in customer support thanks to AI agents; the company later explained that those roles were no longer being backfilled.8 Klarna, the Swedish buy-now-pay-later company, claimed in 2024 that its AI assistant was doing the work of 700 agents, and in May 2025 it started hiring people for customer service again because quality had suffered.9
When you ask companies directly, the picture is calmer. A survey of nearly 6,000 senior managers at firms in the US, the UK, Germany and Australia, published by the National Bureau of Economic Research, found that nine in ten had seen no effect of AI on employment over the past three years. For the next three they expect an average fall of 0.7%.10 The Federal Reserve Bank of New York, which surveys businesses in New York and northern New Jersey every year, found in September that 61% of service firms use AI, up from 40% a year earlier. Only 4% had laid anyone off because of AI. But 15% said they had hired fewer people than they would have without it, and 13% said they had hired more.11
In Italy the word is starting to appear in collective redundancy procedures. In April Nielsen began laying off 43 people at Assago, near Milan, explaining that the production of audience data would move to AI systems. In October the unions also linked to automation the 190 redundancies at CWT, the business travel agency, which however come after two changes of ownership.12
Where the signal is: people who have yet to get in
Mass layoffs are not showing up, but something does show up in the data, and it concerns people looking for work for the first time.
In August 2025 Erik Brynjolfsson, Bharat Chandar and Ruyu Chen of the Stanford Digital Economy Lab analysed payroll records processed by ADP, the largest payroll software provider in the US. In the occupations most exposed to AI, such as software development and customer service, employment of 22 to 25-year-olds had fallen by 13% compared with peers in less exposed jobs.13 On 12 August 2026 they updated the study with data up to June: the gap has grown to 19%, while there is no comparable gap for more experienced workers. According to the data, the mechanism is this: companies are not firing more, they are hiring less. The authors remain cautious. They speak of descriptive signs, the "canaries in the coal mine" of the title, not of proof that AI is the cause, and they note that the gap narrows once education is taken into account.14
Two Harvard economists, Seyed Hosseini and Guy Lichtinger, reach a similar conclusion by a different route. They followed the CVs of 65 million workers at more than 280,000 American companies. Where a company started using generative AI, after six quarters it had about 9% fewer junior employees than companies that had not adopted it, while senior staff numbers did not change. Here too, slower hiring is the main factor. And there is a detail that matters more than the number: the tasks AI can do are disappearing from job adverts for juniors.15
Not everyone reads the data this way. According to the Economic Policy Institute, unemployment among graduates aged 22 to 27 rose from 4.0% in July 2023 to 5.3% in March 2026, but it rose almost as much for peers without a degree, who are far less likely to work in occupations exposed to AI. For the EPI the problem is that American companies are hiring little: the hires rate is back at 2013 and 2014 levels, when the country was still climbing out of the Great Recession.16 Yale's Budget Lab, in its update of 15 September, also finds no clear evidence of labour market disruption linked to AI.17
The two readings are closer than they look. In a market where little hiring is going on, AI gives companies one more reason not to reopen the junior position. Nobody needs to be fired: it is enough not to replace the people who leave. It is the same pattern as Salesforce.
And in Italy
Italy has no public count of announced job cuts with their stated reasons, so the data has to be found elsewhere. The reference estimate of exposure comes from the Bank of Italy, in October 2024: about nine million workers are in occupations highly exposed to AI, mostly in services, with many women and many qualified people. For more than half of them AI would be a help, because it complements the tasks they do. The same study notes, however, that people rarely move between occupations today, and that those who leave an exposed, replaceable job usually end up in a less exposed one that pays less.18
Here the problem for young people comes before AI. In August unemployment among 15 to 24-year-olds was 20.3%, up 0.6 points in a month. Over the same period total employment was 291,000 higher than a year earlier, while among 15 to 24-year-olds it was falling.19 The share of 15 to 29-year-olds who are neither studying nor working has fallen a long way in ten years, from 25.7% in 2015 to 13.3% in 2025, but it remains above the EU average of 11%.20
In a country with these numbers, an AI that takes away beginners' first tasks risks narrowing an entrance that is already narrow.
What I think
I think "AI" has become a respectable label for cuts companies would make anyway. Telling investors you are reducing staff because demand has fallen sounds like a problem; saying you are doing it thanks to AI sounds like a plan. It is not always untrue, but when a number is announced before the technology has done the work, it is a promise more than a fact. And headlines counting 120,000 people "laid off because of AI" frighten people in work without explaining anything to them.
I see the real risk somewhere else, and it is slower and less visible. The tasks AI does best are often the ones through which people used to learn a trade: fixing code, reviewing documents, drafting a first version, answering simple requests. They were dull, but they were the training. If a company hands them to AI and stops hiring juniors, it saves money today and in ten years finds itself without the experienced people who are supposed to check the AI's work. I wrote about this in the piece on who checks the code AI writes: you need people able to judge, and those people are formed by doing.
Every innovation has a hidden price.
Third law of innovation, from Un perché alla volta (translated)
In this case the price may be paid by young people who chose nothing: they simply arrived a few years later. That is why I would ask a company adopting AI whom it will help grow, as well as how much it will save. Innovation is still about people, even when they are the people it has not hired yet.
What companies and young people can do
- HireKeep the entrance openSet a minimum number of junior posts every year, even when AI becomes part of their work. They are tomorrow's experienced staff.
- RedesignGive juniors new tasksNot the ones AI already does, but checking what it produces, with a senior who reviews the work and explains the mistakes.
- Be honestGive the real reasonIf a cut depends on costs or demand, say so. Blaming AI confuses the people who stay and those deciding what to study.
- LearnFor people looking for workUse AI in your own projects and be able to explain how you checked the result. It is a skill an interview can put to the test.
With the next Challenger report, it pays to look at several lines together, not just the AI one. In September, for example, American companies announced 90,787 hires, the lowest figure for that month since 2011.1