The AI Layoff Wave, 2023–2026: How It Moved in Step

In February 2023, some of the world’s largest employers told staff not to use ChatGPT. By mid-2025, several well-known chief executives were telling staff that AI use was no longer optional, that it would count in performance reviews, and that teams had to show why AI could not do a job before hiring for it. From late 2025 to mid-2026, AI was the most-cited reason for announced US job cuts. Then, in the second half of 2026, the US numbers fell while Europe’s rose.

This report follows that sequence step by step, with a source for every number. It is written for people going through a layoff right now, and for anyone who wants the record laid out plainly. I compiled public data from Challenger, Gray & Christmas, Eurofound, the UK Insolvency Service, Indeed Hiring Lab, central-bank and university research, court and regulator records, declassified government documents and major news reporting, as available on 6 November 2026.

On wording: documented means a primary source, or several consistent major outlets, says so. Reported means one outlet or a secondary source says so and the primary was not checked. Unverified means the claim circulates without a reliable source. Vendor surveys and aggregators are flagged as such.

Contents · 10 sections

In one minute #

• The wave moved in step. Public ChatGPT bans clustered in February to May 2023. Approved tools followed over 2023 to 2025. “AI-first” CEO memos came in a burst from April to August 2025. The largest AI-labelled layoff wave ran from Q4 2025 to mid-2026, followed by voluntary-exit programmes.
• The US peaked, then fell. US employers announced 1.21 million job cuts in 2025. In the first half of 2026, AI was cited for 101,743 of them. By October 2026, monthly US cuts were at a three-year low.
• Europe went the other way. Eurofound logged 181,051 announced job losses in the 100 days to 6 November 2026, about double its late-September figure. UK redundancy notifications in September 2026 were 2.1 times the year before.
• The synchronization ran through public channels. CEO memos repeated each other within weeks. Investors rewarded “efficiency.” Peer-reviewed research shows downsizing spreads by imitation. The same consultancies and vendors sold the same playbook, and shared shocks hit everyone at once.
• AI was blamed for more than it measurably did. Oxford Economics found AI cited for only 4.5% of 2025 US cuts, while productivity growth slowed. The clearest measured effect is on young workers: a 13% relative employment drop for 22- to 25-year-olds in the most exposed jobs.
• What the record does not show: any public evidence of a secret agreement between companies to time or size layoffs.
• History matters. States and industries have run documented persuasion campaigns and silenced insiders. The second half of this report sets out that record and the tests you can use to judge the AI-layoff wave yourself.

The timeline, phase by phase #

Phase 1: Bans on public chatbots (February to May 2023) #

JPMorgan restricted staff use of ChatGPT in February 2023. Verizon blocked it from corporate systems while calling AI “integral” to its strategy (HR Brew). Samsung banned generative AI tools on company devices in May 2023 after engineers uploaded source code to ChatGPT (Semafor). Apple, as reported by the Wall Street Journal, and Goldman Sachs, Citigroup and Deutsche Bank were on the same spring lists (Fortune). A BlackBerry survey published in August 2023 found about 75% of organizations implementing or considering bans on work devices, mainly over data security (BlackBerry).

The bans did not stop use. By May 2024, 78% of people using AI at work were bringing their own tools (Microsoft Work Trend Index 2024).

Phase 2: Approved tools (2023 to 2025) #

The fix was an enterprise product or a walled internal platform. ChatGPT Enterprise launched in August 2023 (OpenAI). Orange built an internal AI platform in 2023 that more than 100,000 staff had used by 2026 (TelecomTV). JPMorgan rolled out its in-house LLM Suite in July 2024 (Fortune). Samsung’s global ChatGPT Enterprise rollout was reported only in June 2026 (People Matters). By McKinsey’s 2025 survey, 88% of organizations used AI regularly in at least one function, up from 78% a year earlier (McKinsey). This phase was staggered over about three years.

Phase 3: “AI-first” memos and AI in reviews (April to August 2025) #

This is the most tightly synchronized part of the record.

• Shopify, April 2025: “reflexive AI usage” became a baseline expectation, AI use went into performance reviews, and teams had to show why AI could not do the work before requesting headcount (Anchorage Daily News / Washington Post).
• Duolingo, late April 2025: AI use would count in hiring and reviews, contractors doing automatable work would be phased out, and headcount would come only if a team could not automate (Entrepreneur). Weekly “FrAIdays” followed in June (IT Brew).
• Amazon, 17 June 2025: CEO Andy Jassy wrote that Amazon “will reduce our total corporate workforce as we get efficiency gains from using AI extensively” (TechCrunch).
• Microsoft, late June 2025: a developer-division memo, reported by Business Insider, said “Using AI is no longer optional,” with AI use to feature in performance assessments (Computing).
• Canva, July 2025: about 5,000 staff got a week off regular work to learn AI (HR Grapevine).
• Coinbase, August 2025: CEO Brian Armstrong said he fired some engineers who did not onboard to AI coding tools without a good reason (Fortune).

The shared template had four parts: AI as a baseline, AI measured in reviews, headcount justified against AI, and protected learning time. The third part is the documented link between AI adoption and hiring decisions. Duolingo reportedly later stopped counting AI use in reviews.

Phase 4: Agents and MCP (late 2024 to 2026) #

Anthropic introduced the Model Context Protocol (MCP), a standard for connecting AI models to tools and data, in November 2024 (Anthropic). In December 2025 it was moved to a Linux Foundation fund backed by OpenAI, Google, Microsoft and AWS (Anthropic). Microsoft announced “the Frontier Firm,” built on human-agent teams (Microsoft), and Gartner predicted 40% of enterprise apps would include AI agents by end-2026 (Gartner). Google said about 75% of its new code was AI-generated by April 2026 (MobileSyrup).

Measured deployment lagged. In any single business function, no more than 10% of organizations were scaling agents (McKinsey), and only 31% of developers used agents in 2025 (Stack Overflow).

Phase 5: Peak AI-attributed cuts (Q4 2025 to mid-2026) #

Q4 2025 had the most announced US cuts of any fourth quarter since 2008 (Challenger). October 2025 alone had 153,074, of which 31,039 were attributed to AI, as reported by CFO Dive (CFO Dive).

• Amazon announced about 14,000 corporate cuts in October 2025 and about 16,000 more in January 2026, framed as removing “layers” (AP via KTUL; CRN Asia).
• Lufthansa said in September 2025 it would cut about 4,000 mostly administrative jobs by 2030, citing AI (Skift).
• Meta began laying off about 8,000 people from 20 May 2026, according to Reuters, and changed many manager roles (The Star / Reuters).

In the first half of 2026, AI was cited for 101,743 US cuts, about 23% of the total, and was the top stated reason for four straight months. Coverage named Cloudflare, Snap and Block among the AI-linked cuts (CFO Dive).

Phase 6: Voluntary exits (2025 to 2026) #

Google expanded severance-backed voluntary exits to Canada, France, Germany, Japan and Switzerland in June 2026, excluding its AI and cloud units, according to Reuters (Business Today / Reuters). On 26 October 2026, Microsoft offered voluntary exit to about 60,000 eligible staff, with a deadline in early December (MyNorthwest). Earlier in 2026, Microsoft ran an April retirement offer and then cut jobs in July, as reported by Windows Central (Windows Central).

In Europe, Capgemini planned up to 2,400 French departures through voluntary routes (CFTC). BMW was reported to be running voluntary severance from October 2026 for German staff outside production (SRF). Orange Business temporarily withdrew a plan for about 700 voluntary departures after questions from the labour administration (Le Monde Informatique).

Running underneath: management delayering (since 2023) #

Middle managers made up 32% of layoffs in 2023, up from 20% in 2019, according to Live Data Technologies figures summarized by Cornell (Cornell Catherwood). Middle-manager hiring was down about 43% since 2022, per Revelio Labs (Slashdot on Revelio). GitLab was reported to be removing up to three management layers in 2026 (Final Round AI).

The timeline at a glance #

Phase	When	Key dated facts	Grade
Bans on public chatbots	Feb to May 2023	JPMorgan (Feb), Samsung (May); ~75% implementing or considering bans (Aug 2023)	Documented
Approved tools	2023 to 2025	ChatGPT Enterprise (Aug 2023); JPMorgan LLM Suite (Jul 2024); 88% of firms using AI (2025)	Documented; staggered
AI-first memos, AI in reviews	Apr to Aug 2025	Shopify, Duolingo, Amazon, Microsoft dev division, Canva, Coinbase	Documented; tightly clustered
Agents and MCP	Nov 2024 to 2026	MCP Nov 2024, Linux Foundation Dec 2025; ≤10% scaling agents per function	Documented; uneven
Peak AI-attributed cuts	Q4 2025 to mid-2026	Highest Q4 since 2008; 101,743 AI-cited US cuts in H1 2026	Documented
Voluntary exits	2025 to 2026	Google (Jun 2026); Microsoft ~60,000 (Oct 2026); European voluntary schemes	Documented
Management delayering	2023 onward	Managers 32% of 2023 layoffs; Amazon “layers” (Jan 2026)	Documented

The numbers: the US fell while Europe rose #

United States #

Challenger counted 1,206,374 announced US job cuts in 2025, up 58% from 2024. AI was cited for 54,836. The biggest single reason was federal “DOGE Actions,” at 293,753 (Challenger year-end 2025). AI was the top cited reason in May, June and July 2026 (Challenger May; June; July).

Then the curve turned. October 2026 had 43,994 cuts, the lowest October since 2023 and 71% below October 2025. January to October 2026 was 44% below the same period of 2025. AI was cited for 1,265 cuts in October, the year’s low, and “market and economic conditions” became the top reason for 2026. Hiring plans were up 34% (Challenger October 2026). On Challenger’s figures, AI-cited cuts were about 101,700 in January to June 2026 and only about 19,700 in July to October.

Methods matter. A Business Insider analysis, as reported by Playtech, found AI mentioned in about 8% of 2026 US layoff plans, against Challenger’s 20 to 23% (Playtech.ro).

Europe #

Three independent series point the same way.

• Eurofound’s European Restructuring Monitor, which records announcements of 100 or more job losses, logged 181,051 announced losses across 289 events from 29 July to 6 November 2026. Its 100-day figure to 23 September was about 88,470 (Eurofound ERM). Autos led: Schaeffler reported 3,460 more European cuts in early November (web.de).
• UK HR1 notifications, filed before proposing 20 or more redundancies at one site, covered 50,904 potential redundancies in September 2026, against 24,504 a year earlier (DCA Business Recovery; gov.uk data).
• Indeed job postings in the euro area were 13% lower than a year earlier in October 2026. No European country analysed had positive postings growth, while US postings “recently turned positive.” Software development, marketing and media were weakest (Indeed Hiring Lab).

Measure	United States	Europe (EU and UK)
Announced cuts, 2026	Down 44% Jan to Oct; October a three-year low	ERM 100-day losses roughly doubled Sep to Nov; UK HR1 Sept 2.1x
AI as stated reason	Top reason May to July, then 1,265 in Oct	Rarely cited formally; disputed by unions
Hiring demand	Postings turned positive	Euro-area postings down 13%
Leading sectors	Tech; federal government (2025)	Autos and suppliers, banks, telecoms, IT services

European firms rarely give AI as the formal reason. At Capgemini, a CFTC union official said the market, more than AI, was driving cuts (CFTC). A few cases do cite AI, such as Finnish software firm Relex, with up to 288 cuts (ERM 300290). Eurofound began a dedicated series on AI-mentioning restructuring in July 2026 (Eurofound).

Romania shows the European pattern in sharper form. Experis put its net tech hiring intention at -11% for Q3 2026, last of all countries surveyed, against +35% globally (Undelucram). Industry bodies describe a second wave of IT job losses in 2026, mostly restructurings decided by foreign parent groups (Undelucram; Economica.net).

How synchronization happens #

Many companies cutting at the same time, in the same words, after adopting the same tools, looks orchestrated. The record shows it was coordinated in practice: firms were kept in step by shared, visible channels. Each has its own evidence.

1. The same memos, weeks apart #

Shopify’s CEO posted his memo publicly in April 2025. Within about five months, Duolingo, Amazon, Microsoft’s developer division and Coinbase had issued directives with the same elements (Anchorage Daily News / Washington Post; Inquirer). Press coverage set the memos side by side, making each a reference point for the next.

2. Investors reward “efficiency” #

After Meta declared 2023 its “year of efficiency” and cut about 11,000 jobs, its stock jumped more than 20% after hours on results (Fortune). In February 2024, with headcount down about 22%, it rose more than 20% again (Fortune). A Cleveland Fed working paper found that when a layoff announcement is read as good news, rivals’ share prices rise too, most strongly in technology (Cleveland Fed). One firm’s cuts give its peers a market reason to follow.

3. Peer imitation is a measured effect #

Studying Fortune 100 firms from 1979 to 1994, Art Budros found downsizing spread as a practice, with institutional pressure growing stronger over time while economic necessity faded (Budros, Sociological Forum). A study of Spanish firms found companies copy their industry’s downsizing choices, especially those of leaders (Sánchez-Bueno and Muñoz-Bullón). In Japan, reputable firms resisted at first, then gave way as more joined (“safety in numbers”) (Safety in Numbers). Stanford’s Jeffrey Pfeffer calls the tech layoff wave “social contagion” (Stanford GSB). An IG Metall survey of German works councils found 15% of plants using the downturn to cut jobs even without economic necessity (IG Metall, metall magazine 1/2026).

4. The same consultancies, the same playbook #

Management-fashion research shows how practices spread through consultants and the business press, in upswings marked by “emotionally charged, enthusiastic, and unreasoned discourse” (Abrahamson and Fairchild, ASQ). Consultancies ran the AI playbook on themselves in public. Accenture said in September 2025 it was “exiting” staff it could not reskill, while growing its AI and data specialists to 77,000 (Irish Times). Bloomberg reported McKinsey cut about 200 tech jobs as it shifted roles to AI (ThePrint / Bloomberg).

5. The same vendors and forecasts #

A handful of vendors and analysts published the framing buyers repeated: Microsoft’s “Frontier Firm,” Gartner’s agent forecast and the vendor-backed MCP foundation, all cited above. That gave competing firms the same roadmap at the same time. No source found documents vendor pitches promising specific headcount cuts.

6. Shared shocks #

The Federal Reserve raised rates about 4.5 percentage points from March 2022, after a pandemic hiring boom (Federal Reserve). Meta’s CEO said in 2023 the company had “misjudged the durability of pandemic-era trends” (Anadolu Agency). Hyperscaler AI capital spending for 2026 was estimated at $600 billion to $800 billion (Yahoo Finance). In Europe, Indeed notes a 2026 ECB rate rise and French bond yields near 5%. The stated reason for cuts moved from over-hiring to efficiency to AI, while the behaviour stayed the same.

7. AI-washing: a story that pays #

Once investors reward the “AI efficiency” explanation, each firm has its own reason to use the same words. The term “AI-washing” for blaming financially motivated cuts on AI is attributed to a Forrester report in January 2026 (Fortune). A shared, rewarded narrative synchronizes behaviour on its own.

8. In Europe, the law sets the order #

Directive 98/59/EC requires consultation on avoiding or reducing dismissals before they happen (EUR-Lex). French collective voluntary-departure schemes, German severance programmes agreed with works councils and UK collective consultation typically open with voluntary exits. “Voluntary first, compulsory if take-up falls short” is the standard legal route, so firms follow the same sequence without contacting each other.

What the record does not show #

Big tech firms have colluded on labour before. Apple, Google, Intel, Adobe and others faced a 2010 Justice Department action and a class action over agreements not to poach each other’s staff; the main settlement was $415 million (Wikipedia). That came to light through investigation and court discovery.

For 2023 to 2026, no lawsuit, regulator action, leak or whistleblower has produced evidence of an agreement between companies on the timing or size of layoffs. Firms also diverged: Klarna reversed course on replacing service staff with AI (Bloomberg). Private communications between boards, investors and chief executives are not public, and absence of evidence is not proof of absence. On the public record, the synchronization runs through the open channels above.

What AI actually changed, and what it is blamed for #

Evidence that AI is over-credited #

• Oxford Economics (January 2026): AI was cited for about 55,000 US cuts in 2025, 4.5% of the total; cuts blamed on market conditions were four times larger. Productivity growth slowed. Its economist Ben May said: “We suspect some firms are trying to dress up layoffs as a good news story” (Fortune).
• Sam Altman (February 2026): “there’s some AI washing where people are blaming AI for layoffs that they would otherwise do,” as reported by Built In (Built In).
• Yale Budget Lab: little evidence that generative AI has changed employment since ChatGPT’s release; its director suggested some firms cite AI to mask trade and policy problems (Fortune). July 2026 coverage reported jobs in the most AI-exposed occupations up about 4% (Ohio CPA).
• MIT NANDA (2025): about 95% of organizations saw no measurable profit-and-loss return from generative AI (MIT NANDA). Critics note a narrow six-month definition of success (Zirous).

AI went from a small stated reason in 2025 to the top stated reason in mid-2026, while aggregate labour data barely moved. A stated-reason series that jumps while outcomes stay flat fits a change in the story better than a change in the technology. That fits AI-washing in general without proving it for any single firm.

What AI is measurably doing #

• Young workers first. Stanford researchers Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, using ADP payroll data, found a 13% relative employment decline for 22- to 25-year-olds in the most AI-exposed occupations. Older workers in the same jobs were stable or growing (Stanford SIEPR).
• Small but rising. In New York Fed surveys, 4% of service firms laid off workers because of AI in 2026, up from 1% a year earlier; no manufacturers did. Retraining was the main response (Liberty Street Economics).
• A forecast, not a measurement. Goldman Sachs estimates AI could displace 6 to 7% of the US workforce over the adoption period, mostly temporarily (Goldman Sachs).

Reversals and rehiring #

Klarna froze most hiring from December 2023 to replace labour with AI (Fortune). In May 2025, its CEO told Bloomberg that cost had been “a too predominant evaluation factor,” producing “lower quality,” and Klarna began hiring human agents again (Bloomberg). Ford and Commonwealth Bank of Australia were reported to have reversed AI-linked cuts after automated systems failed (Washington Times; Quartz). Surveys by Orgvue (32%) and Robert Half (about 29%) found that many organizations that cut for AI later rehired, and Forrester found 55% of employers regret AI layoffs (HCAMag). These are vendor surveys with commercial interests and no national count behind them, so treat them as indicators.

The documented history of persuasion and silencing #

Some readers will ask whether a wave this synchronized is social engineering. That question has a factual history. Governments and industries have run deliberate persuasion programmes, and the proof sits in congressional reports, released files, court rulings and major investigations.

Mass persuasion as a trade #

The US Committee on Public Information (1917 to 1919), the Creel Committee, was the government’s official wartime propaganda agency; its final report is public (archive.org). Edward Bernays, who worked in its orbit, carried its methods into corporate PR. In Propaganda (1928) he wrote that “the conscious and intelligent manipulation of the organized habits and opinions of the masses is an important element in democratic society” (archive.org). That is a practitioner advertising his trade: evidence of intent and method, not of effect.

What the Church Committee found #

The Senate’s Church Committee investigated the intelligence agencies in 1975 and 1976 (Senate Intelligence Committee; Book II; Book III).

• COINTELPRO. Book II records that “under the cloak of secrecy, the FBI instituted its COINTELPRO operations to ‘disrupt’ and ‘neutralize’ ‘subversives’.” Book III details the methods, including propaganda and anonymous letters to targets’ employers and newspapers (FBI Vault). It is the clearest documented case of a state agency deliberately trying to cost critics their jobs.
• The CIA and the press. A committee staff document released in 2018 under the JFK Records Act says that of the people in CIA media relationships, “little over one third are journalists; another third are CIA agents operating under media cover” (National Archives). The CIA then adopted a 1976 policy of no secret paid relationships with accredited US journalists. Two caveats: in the CIA’s released “Family Jewels,” Project MOCKINGBIRD was a 1963 wiretap of two journalists, not a media-control network (CIA Reading Room); and the popular “400 journalists” figure comes from a 1977 Rolling Stone article, not an official finding.

MKUltra: experiments on people, then destroyed records #

The 1977 Senate hearing on Project MKULTRA documents CIA drug and behavioural experiments on human subjects, some unwitting. It records that in January 1973 MKUltra records were destroyed on the verbal orders of Sidney Gottlieb, carrying out an order from then-Director Richard Helms (archive.org). It was experimentation on individuals, not a mass-media campaign.

Online persona campaigns #

• Operation Earnest Voice: US Central Command contracted in 2010 to 2011 for “persona management” software to run fake identities on foreign-language social media (Wikipedia; The Black Vault).
• GCHQ’s JTRIG: documents leaked in 2014 show a British unit trained in fake personas, “false flag” posts and discrediting operations (Common Dreams). They show capability and doctrine; specific operations are only partly documented.
• The Pentagon and Sinovac: Reuters reported in June 2024 that the US military ran fake accounts posing as Filipinos to cast doubt on China’s Sinovac vaccine from 2020 to mid-2021, which a senior Defense Department official acknowledged (Reuters).

Corporate campaigns found in court and in archives #

• Tobacco: a 1969 Brown and Williamson memo said “Doubt is our product.” Millions of pages of industry documents were released through litigation (UCSF Industry Documents). In United States v. Philip Morris (2006), a federal court found the major tobacco companies violated the RICO Act by deceiving the public about smoking’s risks (opinion).
• Sugar: a 2016 JAMA Internal Medicine analysis of internal documents found the Sugar Research Foundation paid Harvard scientists in the 1960s for a review that played down sugar’s role in heart disease (PubMed).
• Climate: a 2023 study in Science found ExxonMobil’s own scientists modelled global warming accurately (PubMed).
• Cambridge Analytica: the US Federal Trade Commission found in 2019 that the firm deceived consumers about collecting Facebook data (FTC). The UK Information Commissioner found its targeting less effective and novel than claimed (ICO).

Government pressure on platforms: what the court held #

In Murthy v. Missouri (June 2024), plaintiffs argued federal officials coerced platforms into removing content. The Supreme Court ruled 6 to 3 that they lacked standing and did not decide whether the conduct was unconstitutional. Justice Barrett wrote that many of the lower court’s factual findings “appear to be clearly erroneous.” Justice Alito, dissenting with Justices Thomas and Gorsuch, described a serious coercion campaign (Supreme Court). Government contact with platforms is documented; whether it crossed the constitutional line is legally undecided.

Silencing insiders, 2023 to 2025 #

• OpenAI, 2024: Vox reported that departing staff could lose vested equity if they did not sign non-disparagement agreements; OpenAI said it removed those terms. Whistleblowers then asked the SEC to investigate, alleging the agreements discouraged reporting to regulators (TechCrunch; Kohn, Kohn and Colapinto). No SEC outcome has been found.
• Meta, 2025: after performance layoffs of about 5% of staff, Meta fired about 20 employees it said had leaked confidential information, including a Zuckerberg memo (Entrepreneur).
• US severance gag clauses: in McLaren Macomb (2023), the National Labor Relations Board held that broad confidentiality and non-disparagement clauses in severance agreements unlawfully restrict employees’ rights (NLRB). In February 2025, the NLRB’s acting general counsel rescinded the guidance applying it broadly; the decision itself stands until the Board overrules it (Mintz).

Coverage of AI layoffs, including sceptical “AI-washing” reporting, is abundant. No documented case was found of a journalist fired, sued or removed from a platform for reporting on them. The documented silencing is of employees, through contracts.

How to judge it yourself #

Every documented programme above left the same traces: a named operator, a defined target, a paper trail (contracts, memos, budgets), evidence of coordination between the parties, eventual exposure (by Congress, courts, regulators, leaks or reporters), and often measures to keep insiders quiet. Here is how the AI-layoff wave stands on the public record as of November 2026:

Test	AI-layoff wave, 2023 to 2026
Public memos with a shared template	Documented (2025)
Investor rewards for cuts	Documented (Meta 2023 and 2024; Cleveland Fed rival effects)
Shared advisers and vendors	Documented (Accenture, McKinsey; Microsoft, Gartner, MCP foundation)
A story larger than the measured effect	Documented (Oxford Economics, Yale Budget Lab, Altman)
Insider gagging	Documented (OpenAI exit terms, Meta leak firings, NLRB guidance rescinded)
Communications or agreement between firms	Not found
A named operator running a cross-company campaign	Not found
Court, regulator or congressional finding	None

The record shows a wave synchronized and amplified through open channels, a story bigger than the facts, and contracts that keep insiders quiet. It does not show a central hand. Whether that adds up to social engineering is your judgement to make.

If you are affected #

This is general information, not legal or financial advice. Confirm anything you act on with a lawyer, an accountant, or your union or employee representatives.

Your rights under EU law #

Directive 98/59/EC sets the EU minimum for collective redundancies (EUR-Lex). It applies when an employer plans dismissals above set thresholds within 30 days, for example at least 10 in a firm of 21 to 99 staff, 10% in a firm of 100 to 299, or 30 in a firm of 300 or more. The employer must:

• consult workers’ representatives “in good time” on ways to avoid or reduce dismissals and soften their effects, including retraining;
• give them in writing the reasons, numbers and categories affected, the period, and the selection criteria;
• notify the public authority, and wait at least 30 days before dismissals take effect.

These duties apply even when a parent company abroad takes the decision.

Romania: what the law gives you #

Not legal or tax advice. Confirm with a lawyer and an accountant before you sign anything.

Consultation (Labour Code art. 68 to 74). The employer must state in writing the reasons, numbers, categories, selection criteria, measures to limit dismissals, compensation and dates. Representatives have 10 calendar days for counter-proposals; the employer answers in writing within 5. The employer must notify the labour inspectorate (ITM) and employment agency (AJOFM) at least 30 calendar days before issuing decisions. This applies in full when a foreign parent decides (Labour Code, Rubinian).

Notice and rehiring. At least 20 working days’ notice (WageIndicator). For 45 days after dismissal, you have priority to be rehired, without exam or probation, if your post is re-created (art. 74).

Selection criteria. Courts have annulled decisions that did not state the criteria and why the employee was chosen (legeaz.net). Ask for the criteria and your own score in writing.

No statutory severance. It exists only if your contract, collective agreement or an internal policy provides it, or you negotiate it. Romanian IT mutual-agreement packages of 10 months’ salary or more, up to 14 to 16 at some firms, were reported in 2026 (Undelucram).

Tax depends on the label. Collective-dismissal compensation paid under your contract or collective agreement is taxed as salary: 10% income tax, 25% CAS (pension), 10% CASS (health). An amicable payment with no contractual or legal basis is “other income”: 10% tax withheld, no CAS, and you self-declare CASS if such income exceeds six minimum gross wages (Fiscalitatea, May 2026). The net difference can be large, and the salary route may count toward your pension. Ask an accountant to model both.

Unemployment benefit (Law 76/2002). RON 495 (75% of the RON 660 reference indicator) plus 3%, 5%, 7% or 10% of your average gross salary for 3, 5, 10 or 20+ years of contributions, for 6, 9 or 12 months, with 10% CASS withheld. You need 12 months of contributions in the last 24, a job loss not attributable to you, and AJOFM registration; it stops if you leave the country for over three months (Playtech). Check with AJOFM whether a mutual-termination agreement affects eligibility before you sign one.

Non-compete. Enforceable only if it names activities, area and duration and pays at least 50% of your average gross salary (art. 21). Ask for a written waiver or confirmation of payment.

EU Globalisation Adjustment Fund. Under Regulation (EU) 2026/1139, the EGF can fund retraining before layoffs for 200+ workers in one enterprise, but only the employer and the state can apply, within 14 weeks of notification (Gruia Dufaut). Ask in consultation whether an application is planned. AJOFM also offers free training to registered jobseekers (ANOFM).

Deadline to contest. A 45-day deadline is commonly cited but unverified here. Confirm it with a lawyer early.

Anywhere, you can usually ask for these in a settlement: release from working notice, payment for unused leave, a pro-rata bonus, treatment of shares, an agreed reference and announcement wording, and an outplacement or training budget.

UK and US pointers #

• UK: employers proposing 20 or more redundancies at one establishment must file an HR1 and consult collectively, for at least 45 days where 100 or more are proposed (gov.uk). Consultation usually opens with a call for volunteers.
• US: read severance agreements for confidentiality and non-disparagement clauses. McLaren Macomb still limits broad gag clauses for non-supervisory employees (Mintz), and SEC rules bar agreements that impede reporting to the SEC (Kohn, Kohn and Colapinto).

Who is still hiring #

• Senior over junior. In the US, 71% of the past year’s rise in software postings came from senior roles (Indeed Hiring Lab).
• Security, infrastructure and AI engineering. Romanian industry reporting names cyber-security, AI and machine-learning engineering, senior developers and senior DevOps as still in demand, while juniors and generalists take the losses (Undelucram).
• Regulation-driven work. The NIS2 cybersecurity directive is in force in Romania through Emergency Ordinance 155/2024, covering energy, banking, health, telecoms and public administration (Clifford Chance). The AI Act’s high-risk obligations were agreed to move to December 2027 and August 2028 (JDSupra), opening a window for compliance work.
• Plan for a long search. People often return at a lower level than they left. Treat severance and benefits as runway, and budget for months, not weeks.

Look after yourself #

A layoff is a loss, even when nobody made it personal. Talk to someone: a friend, family, your doctor, a counsellor. You do not have to carry it alone.

Questions people ask #

Did companies secretly agree to time the AI layoffs? #

There is no public evidence of that. No lawsuit, regulator action, leak or whistleblower has shown an agreement between firms. What is documented is synchronization through open channels: copied memos, investor rewards, measured peer imitation, shared consultancies and vendors, and shared economic shocks. Big tech has colluded on labour before, in the no-poach case, which surfaced through investigation and court discovery.

Is AI actually taking jobs? #

Some, mostly at the entry level so far. Stanford found a 13% relative employment drop for 22- to 25-year-olds in the most exposed jobs, and the New York Fed found 4% of service firms cutting because of AI in 2026. In aggregate, Yale Budget Lab and Oxford Economics find little displacement. AI is blamed for more than it has measurably done.

Why did Europe get worse in late 2026 while the US improved? #

Mainly the economy: the auto-sector crisis, an ECB rate rise and French bond-market stress. Eurofound’s announced losses roughly doubled between late September and early November, while US cuts were down 44% on the year and US postings turned positive.

If an employer offers voluntary exits, do compulsory cuts follow? #

Not always, but there are precedents, such as Microsoft’s April 2026 retirement offer followed by July cuts, as reported. In Europe, voluntary schemes are often the legally expected first step. Ask your representatives what happens if take-up falls short, in writing if you can.

Can a severance agreement stop you from talking? #

It may try. In the US, McLaren Macomb and SEC rules limit how far it can go. In the EU and Romania, read the wording, ask a lawyer what it covers, and negotiate narrower language before signing.

Sources #

Timeline: bans and approved tools

1. HR Brew, “These companies have banned ChatGPT in the office”: https://hr-brew.com/stories/2023/05/11/these-companies-have-banned-chatgpt-in-the-office
2. Semafor, companies banning ChatGPT: https://www.semafor.com/article/05/19/2023/chatgpt-companies-banning-workers
3. Fortune, ChatGPT banned at Apple, Goldman and others: https://fortune.com/2023/05/19/chatgpt-banned-workplace-apple-goldman-risk-privacy
4. BlackBerry, why companies ban ChatGPT: https://blogs.blackberry.com/en/2023/08/why-companies-ban-chatgpt-ai
5. Microsoft and LinkedIn, 2024 Work Trend Index: https://blogs.microsoft.com/blog/2024/05/08/microsoft-and-linkedin-release-the-2024-work-trend-index-on-the-state-of-ai-at-work/
6. OpenAI, Introducing ChatGPT Enterprise: https://openai.com/index/introducing-chatgpt-enterprise/
7. TelecomTV, Orange and AI: https://www.telecomtv.com/content/ai/how-orange-is-putting-ai-at-the-heart-of-its-new-strategy-55105/
8. Fortune, JPMorgan LLM Suite: https://fortune.com/2024/07/26/jpmorgan-chatbot-llm-suite-chatgpt-ai-machine-learning-wall-street-banking-finance
9. People Matters, Samsung ChatGPT and Codex rollout: https://www.peoplematters.in/news/ai-and-emerging-tech/samsung-rolls-out-chatgpt-and-codex-to-employees-worldwide-50429
10. McKinsey, The State of AI: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

Timeline: AI-first memos and agents

11. Anchorage Daily News / Washington Post, “No AI, no job”: https://www.adn.com/nation-world/2025/06/03/no-ai-no-job-these-companies-are-requiring-workers-to-use-the-tech/
12. Philadelphia Inquirer, Duolingo and Shopify: https://www.inquirer.com/business/ai-artificial-intelligence-work-jobs-duolingo-shopify-20250705.html
13. Entrepreneur, Duolingo performance reviews: https://www.entrepreneur.com/business-news/duolingos-ceo-changing-how-he-measures-employee-performance-backlash
14. IT Brew, AI-first at Duolingo: https://www.itbrew.com/stories/2025/08/26/what-ai-first-looks-like-at-duolingo
15. TechCrunch, Amazon expects to reduce corporate jobs due to AI: https://techcrunch.com/2025/06/17/amazon-expects-to-reduce-corporate-jobs-due-to-ai
16. Computing, Microsoft mandates AI use: https://www.computing.co.uk/news/2025/ai/microsoft-mandates-ai-use-by-employees
17. HR Grapevine, Canva AI week: https://www.hrgrapevine.com/us/content/article/2025-07-09-canva-gives-5000-employees-the-week-off-to-learn-ai
18. Fortune, Coinbase AI coding mandate: https://fortune.com/2025/08/25/coinbase-ceo-brian-armstrong-ai-coding-assistants-mandate-tech
19. Anthropic, Introducing the Model Context Protocol: https://www.anthropic.com/news/model-context-protocol
20. Anthropic, donating MCP and the Agentic AI Foundation: https://anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation
21. Microsoft, 2025 Work Trend Index: https://blogs.microsoft.com/blog/2025/04/23/the-2025-annual-work-trend-index-the-frontier-firm-is-born/
22. Gartner, 40% of enterprise apps with AI agents: https://www.gartner.com/en/newsroom/press-releases/2025-08-26-gartner-predicts-40-percent-of-enterprise-apps-will-feature-task-specific-ai-agents-by-2026-up-from-less-than-5-percent-in-2025.
23. MobileSyrup, AI generating most of Google’s code: https://mobilesyrup.com/2026/04/24/ai-is-generating-most-of-googles-code/
24. Stack Overflow, 2025 Developer Survey: https://stackoverflow.co/company/press/archive/stack-overflow-2025-developer-survey/

Timeline: cuts, voluntary exits, delayering

25. CFO Dive, October 2025 job-cut surge: https://www.cfodive.com/news/cost-cutting-ai-drive-october-job-cut-surge-challenger/804935/
26. AP via KTUL, Amazon corporate cuts: https://ktul.com/news/nation-world/amazon-cuts-30000-corporate-jobs-report-tech-seattle-workers-employment-layoffs-hiring
27. CRN Asia, Amazon confirms 16,000: https://www.crnasia.com/news-network/2026/amazon-confirms-16
28. Skift, Lufthansa to cut 4,000 jobs: https://skift.com/2025/09/29/lufthansa-to-cut-4000-jobs-by-2030-as-it-looks-to-ai/
29. The Star / Reuters, Meta May 2026 layoffs: https://www.thestar.com.my/tech/tech-news/2026/05/19/exclusive-meta-lays-out-plans-for-may-20-layoffs-restructuring-internal-document-says
30. CFO Dive, H1 2026 Challenger data: https://www.cfodive.com/news/tech-layoffs-surge-83percent-h1-2026-challenger-ai-disruption/824260/
31. Business Today / Reuters, Google voluntary buyouts: https://www.businesstoday.in/technology/news/story/google-trims-headcount-the-soft-way-offering-voluntary-buyouts-instead-of-layoffs-565713-2026-11-04
32. MyNorthwest, Microsoft buyouts: https://mynorthwest.com/money/microsoft-offers-buyouts/4290545
33. Windows Central, Microsoft voluntary exit: https://www.windowscentral.com/microsoft/microsoft-extends-voluntary-exit-program-to-60-000-employees-are-layoffs-around-the-corner
34. CFTC, Capgemini: https://www.cftc.fr/actualites/suppressions-de-postes-a-capgemini-on-veillera-a-ce-que-les-departs-soient-volontaires-pas-subis
35. SRF, BMW job cuts: https://www.srf.ch/news/wirtschaft/abbau-bei-autokonzern-jetzt-auch-bmw-konzern-soll-tausende-stellen-streichen
36. Le Monde Informatique, Orange Business departure plan: https://www.lemondeinformatique.fr/actualites/lire-le-plan-de-departs-d-orange-business-pousse-sur-le-bureau-du-ministere-du-travail-91981.html
37. Cornell Catherwood, middle managers and layoffs: https://catherwood.library.cornell.edu/wit/middle-managers-have-been-a-major-target-for-layoffs-in-the-past-year
38. Slashdot on Revelio Labs, middle-manager hiring: https://slashdot.org/story/24/12/03/1534254/middle-manager-hiring-has-plunged
39. Final Round AI, GitLab restructuring: https://www.finalroundai.com/blog/gitlab-ai-restructuring-2026

The numbers

40. Challenger, year-end 2025 report: https://www.challengergray.com/blog/2025-year-end-challenger-report-highest-q4-layoffs-since-2008-lowest-ytd-hiring-since-2010/
41. Challenger Report, May 2026: https://www.challengergray.com/blog/
42. Challenger Report, June 2026: https://www.challengergray.com/blog/
43. Challenger Report, July 2026: https://www.challengergray.com/blog/
44. Challenger, October 2026 report: https://www.challengergray.com/blog/october-layoffs-at-three-year-low-seasonal-hiring-plans-up-26/
45. Playtech.ro, IT layoffs and AI share: https://playtech.ro/2026/it-istii-candva-rasfatatii-sortii-acum-ingroasa-randurile-somerilor-in-romania-peste-2-500-de-posturi-taiate-insa-fenomenul-atinge-apogeul-in-sua-cu-aproximativ-140-000-de-concedieri-totul-se/
46. Eurofound, European Restructuring Monitor: https://apps.eurofound.europa.eu/restructuring-events
47. Eurofound ERM factsheet 300290 (Relex): https://apps.eurofound.europa.eu/restructuring-events/detail/300290
48. Eurofound, restructuring mentioning AI data series: https://www.eurofound.europa.eu/en/surveys-and-data/data-catalogue/announced-net-employment-effects-of-restructuring-mentioning-ai-four-month-moving-average-eu-27-2017-2026
49. web.de, Schaeffler job cuts: https://web.de/magazine/wirtschaft/schaeffler-schliesst-werke-baut-tausende-jobs-42904574
50. DCA Business Recovery, UK HR1 notifications: https://dcabr.co.uk/news/uk-potential-redundancies-more-than-double-as-hr1-notifications-jump
51. gov.uk, HR1 advance notification data: https://www.gov.uk/government/publications/publication-of-data-on-advanced-notification-of-redundancy-scheme
52. Indeed Hiring Lab, European Labour Market Snapshot, October 2026: https://hiringlab.indeed.com/uk/blog/2026/10/26/european-labour-market-snapshot-october-2026/
53. Undelucram, Romania tech hiring at -11%: https://www.undelucram.ro/en/cumlucram/romania-tech-hiring-hits-11-as-global-rate-climbs-to-35-2404
54. Undelucram, Romanian IT sector in 2026: https://www.undelucram.ro/en/cumlucram/romanian-it-sector-in-2026-10000-jobs-lost-and-a-second-wave-under-way-2419
55. Economica.net, Cluj layoffs: https://www.economica.net/un-nou-val-de-concedieri-in-it-aproape-400-de-persoane-vor-fi-disponibilizate-la-cluj-napoca_987607.html

How synchronization happens

56. Fortune, Meta “year of efficiency”: https://fortune.com/2023/03/14/meta-layoffs-10000-mark-zuckerberg-year-of-efficiency-hiring-freeze
57. Fortune, Meta stock after year of efficiency (Feb 2024): https://dc.fortune.com/2024/02/02/meta-stock-earnings-zuckerberg-year-of-efficiency-layoffs-dividend/
58. Cleveland Fed, rival equity prices and mass layoff announcements: https://www.clevelandfed.org/-/media/project/clevelandfedtenant/clevelandfedsite/publications/working-papers/2018/wp-1610r-rival-growth-prospects-and-equity-prices-evidence-from-mass-layoffs-announcements-pdf.pdf
59. Budros, Sociological Forum: https://doi.org/10.1023/A:1016093330881
60. Sánchez-Bueno and Muñoz-Bullón, institutional determinants of downsizing: https://researchportal.uc3m.es/display/act417856
61. Safety in Numbers, downsizing in Japan: https://scinapse.io/papers/2071784820
62. Stanford GSB, Pfeffer on copycat layoffs: https://www.gsb.stanford.edu/insights/why-copycat-layoffs-wont-help-tech-companies-or-their-employees
63. IG Metall, metall magazine 1/2026: https://www.igmetall.de/
64. Abrahamson and Fairchild, Administrative Science Quarterly: https://journals.sagepub.com/doi/10.2307/2667053
65. Irish Times, Accenture to exit staff: https://www.irishtimes.com/business/2025/09/26/accenture-to-exit-staff-who-cannot-be-retrained-for-age-of-ai
66. ThePrint / Bloomberg, McKinsey cuts tech jobs: https://theprint.in/business/mckinsey-cuts-about-200-tech-jobs-as-it-shifts-more-roles-to-ai/2793389
67. Federal Reserve, open market operations (rate history): https://www.federalreserve.gov/monetarypolicy/openmarket.htm
68. Anadolu Agency, why tech companies are laying off: https://aa.com.tr/en/economy/why-are-tech-companies-laying-off-workers/2805267
69. Yahoo Finance, hyperscaler 2026 capex: https://finance.yahoo.com/sectors/technology/articles/hyperscalers-hit-700-billion-2026-111243744.html
70. EUR-Lex, Directive 98/59/EC: https://eur-lex.europa.eu/eli/dir/1998/59/oj
71. Wikipedia, High-Tech Employee Antitrust Litigation: https://en.wikipedia.org/wiki/High-Tech_Employee_Antitrust_Litigation

What AI changed vs what it is blamed for

72. Fortune, Oxford Economics on AI layoffs: https://fortune.com/2026/01/07/ai-layoffs-convenient-corporate-fiction-true-false-oxford-economics-productivity/
73. Built In, AI-washing layoffs: https://builtin.com/articles/ai-washing-layoffs
74. Fortune, Yale Budget Lab: https://fortune.com/2026/02/02/ai-labor-market-yale-budget-lab-ai-washing/
75. Ohio CPA, AI-exposed occupations up 4%: https://ohiocpa.com/get-involved/news/2026/07/24/report–jobs-in-occupations-most-exposed-to-ai-increased-by-4
76. MIT NANDA, “The GenAI Divide” (2025): https://nanda.media.mit.edu/
77. Zirous, critique of the MIT 95% figure: https://www.zirous.com/insights/a-clearer-view-of-mits-genai-data-what-95-of-genai-projects-fail-really-means/
78. Stanford SIEPR, “Canaries in the Coal Mine”: https://siepr.stanford.edu/publications/working-paper/canaries-coal-mine-six-facts-about-recent-employment-effects-artificial
79. New York Fed, Liberty Street Economics: https://libertystreeteconomics.newyorkfed.org/2026/09/businesses-are-using-ai-to-transform-work-not-cut-jobs/
80. Goldman Sachs, AI and the global workforce: https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-global-workforce
81. Indeed Hiring Lab, labor market tilting toward seniority: https://hiringlab.indeed.com/2026/07/23/the-labor-market-is-tilting-toward-seniority/
82. Fortune, Klarna stopped hiring: https://fortune.com/2024/12/12/klarna-stopped-all-hiring-replace-workers-with-ai
83. Bloomberg, Klarna turns from AI to real-person customer service: https://www.bloomberg.com/news/articles/2025-05-08/klarna-turns-from-ai-to-real-person-customer-service
84. Washington Times, AI layoff reversals: https://www.washingtontimes.com/news/2026/mar/10/ai-layoff-reversal-companies-rehire-customer-roles-eliminated/
85. Quartz, companies rehiring after AI layoffs: https://qz.com/companies-rehiring-workers-ai-layoffs-automation-070126
86. HCAMag, rehiring after AI-driven cuts: https://www.hcamag.com/us/specialization/hr-technology/businesses-rush-to-rehire-staff-after-regretted-ai-driven-cuts/568292

Persuasion and silencing

87. Complete Report of the Committee on Public Information, 1917-1919 (archive.org): https://archive.org/search?query=%22Committee+on+Public+Information%22+complete+report
88. Edward Bernays, Propaganda (1928) (archive.org): https://archive.org/details/edward-bernays-propaganda-1928
89. US Senate Select Committee on Intelligence, intelligence-related commissions: https://www.intelligence.senate.gov/resources/intelligence-related-commissions
90. Church Committee, Book II (AARC): https://www.aarclibrary.org/publib/church/reports/book2/contents.htm
91. Church Committee, Book III (AARC): https://www.aarclibrary.org/publib/church/reports/book3/contents.htm
92. FBI Vault, COINTELPRO: https://vault.fbi.gov/cointel-pro
93. National Archives, JFK Assassination Records: https://www.archives.gov/research/jfk
94. CIA Reading Room, Family Jewels: https://www.cia.gov/readingroom/collection/family-jewels
95. Project MKULTRA joint Senate hearing, 1977 (archive.org): https://archive.org/details/projectmkultraci0000unit
96. Wikipedia, Operation Earnest Voice: https://en.wikipedia.org/wiki/Operation_Earnest_Voice
97. The Black Vault, Operation Earnest Voice records: https://www.theblackvault.com/documentarchive/operation-earnest-voice/
98. Common Dreams, GCHQ JTRIG documents: https://commondreams.org/news/2014/02/07/gchqs-dirty-tricks-revealed-false-flags-virus-attacks-and-honey-traps
99. Reuters, Pentagon anti-vax campaign: https://www.reuters.com/investigates/special-report/usa-covid-propaganda/
100. UCSF Industry Documents Library, tobacco: https://www.industrydocuments.ucsf.edu/tobacco/
101. United States v. Philip Morris, final opinion (2006): https://www.publichealthlawcenter.org/sites/default/files/resources/doj-final-opinion.pdf
102. Kearns, Schmidt and Glantz, JAMA Internal Medicine 2016 (PubMed): https://pubmed.ncbi.nlm.nih.gov/27617709/
103. Supran, Rahmstorf and Oreskes, Science 2023 (PubMed): https://pubmed.ncbi.nlm.nih.gov/36634176/
104. FTC, opinion and order against Cambridge Analytica: https://www.ftc.gov/news-events/news/press-releases/2019/12/ftc-issues-opinion-order-against-cambridge-analytica-deceiving-consumers-about-collection-facebook
105. UK ICO, investigation into data analytics for political purposes: https://ico.org.uk/action-weve-taken/investigation-into-data-analytics-for-political-purposes/
106. Supreme Court, Murthy v. Missouri opinion: https://www.supremecourt.gov/opinions/23pdf/23-411_3dq3.pdf
107. TechCrunch, whistleblowers accuse OpenAI over NDAs: https://techcrunch.com/2024/07/13/whistleblowers-accuse-openai-of-illegally-restrictive-ndas
108. Kohn, Kohn and Colapinto, SEC whistleblower complaint on OpenAI: https://kkc.com/media/leaked-sec-whistleblower-complaint-challenges-openai-on-illegal-non-disclosure-agreements/
109. Entrepreneur, Meta fires 20 employees for leaking: https://www.entrepreneur.com/business-news/meta-fires-20-employees-for-leaking-information-to-the-press/487718
110. NLRB, McLaren Macomb decision: https://www.nlrb.gov/news-outreach/news-story/board-rules-that-employers-may-not-offer-severance-agreements-requiring
111. Mintz, NLRB moves away from Biden-era approach to severance agreements: https://www.mintz.com/insights-center/viewpoints/2226/2025-03-13-nlrb-moves-away-biden-era-approach-severance-agreements

If you are affected

112. Rubinian, Romanian Labour Code art. 68-74: https://www.rubinian.com/codul-muncii-titlul-2-capitolul-5-sectiunea-5
113. Rubinian, Labour Code art. 74: https://www.rubinian.com/codul-muncii-titlul-2-capitolul-5-sectiunea-5-articolul-74
114. WageIndicator Romania, notice and compensation: https://wageindicator.org/ro-ro/munca-in-romania/legislatia-muncii/contracte-si-concedieri/cerinta-aviz-si-plati-compensatorii
115. legeaz.net, case law on missing reasons in collective dismissal: https://legeaz.net/spete-dreptul-muncii/concediere-colectiva-neindicarea-motivelor-3qd
116. Fiscalitatea, tax treatment of compensation payments (May 2026): https://www.fiscalitatea.ro/plati-compensatorii-catre-un-fost-salariat-tratament-fiscal-contabil-inregistrari-declaratii-24850/
117. Playtech, unemployment benefit in 2026: https://playtech.ro/2026/cum-se-calculeaza-ajutorul-de-somaj-in-2026-cati-bani-poti-primi-daca-ramai-fara-loc-de-munca/
118. legeaz.net, Labour Code art. 21 (non-compete): https://legeaz.net/legea-53-2003-codul-muncii/art-21
119. Gruia Dufaut, Regulation (EU) 2026/1139 and the EGF: https://www.gruiadufaut.com/en/posts/regulation-eu-20261139-eu-support-for-workers-affected-by-imminent-job-displacement
120. ANOFM, vocational training communiqué: https://www.anofm.ro/
121. Clifford Chance, NIS2 in Romania: https://www.cliffordchance.com/briefings/2025/12/navigating-nis2-cybersecurity-compliance-obligations-for-romanian-companies.html
122. Hogan Lovells via JDSupra, AI Act high-risk delay: https://www.jdsupra.com/legalnews/eu-legislators-agree-to-delay-for-high-5617765/

Facts first #

The record from 2023 to 2026 is clear on several points. Employers banned public chatbots, then approved their own. Chief executives copied each other’s AI memos within weeks. Investors paid for cuts. The same consultancies and vendors sold the same story. AI was cited for far more cuts than it can be shown to have caused, while its measurable effect fell hardest on the youngest workers. The US wave crested in mid-2026 just as Europe’s began to rise. Insiders who might have spoken were bound by contracts or fired for leaks. No public evidence shows a secret pact. The sources are listed above, so you can check each fact and judge the pattern for yourself.