takeoff.watch

A million workers strike over AI or automation

Will a strike or coordinated general strike of at least 1 million workers take place in any country, with the organising unions naming AI or automation as a principal grievance?

Chance by Q4 2031

5%

Cumulative probability

0%25%50%75%100%
  • Latest forecast

Forecast history

Forecast made on 13 Oct 2026

Latest of 5 forecasts · Methodology v2.1

Model reasoning

Forecast made 13 Oct 2026 · Methodology v2.1

There is no recent precedent for a strike of a million or more workers called mainly over artificial intelligence or automation. Today's strikes over these technologies involve tens of thousands of workers at most. The giant national strikes in India, France and Spain concern labour laws, wages and housing, and news agencies rarely confirm turnouts above a million themselves. The forecast would rise sharply if a large national union federation called a general strike centred on job losses from artificial intelligence.

Current status (as of 13 Oct 2026)

Nothing qualifies yet, and no qualifying action is scheduled. To resolve YES, three things must all line up: (a) a single strike call with at least 1M strikers in one country; (b) that turnout confirmed by an official source or by Reuters/AP/AFP in their own words; and (c) the organising union or coalition naming AI or automation as a principal grievance in its common call.

  • AI strikes are spreading but are still small. Hyundai Motor's union held its first full strike in ten years (Aug 2026, about 40,000 workers), with job guarantees against AI and humanoid robots among the main issues. Italy's SLC‑CGIL, FISTel‑CISL and UILCOM ran a national call-centre strike over AI from 13 to 27 July 2026. Journalists (France, Seattle Times) and doctors (Allina) have also struck partly over AI. The largest strike ever with AI as a central issue is still SAG‑AFTRA's in 2023 (about 160,000). That is roughly an order of magnitude below the threshold.
  • The mass strikes this year and next are about other things. India's 12 Feb 2026 general strike was against the labour codes and the US trade deal. The CTU–SKM declaration of 29 July 2026 lists about 20 demands and none concerns AI; AI came up only in speeches ("technological advances, including AI, reduce human hardship"). Spain's 11 Nov 2026 general strike is about housing and wages. The French intersyndicale's call for 5 Nov 2026 is about wages, pensions, public services and industry, with only a passing reference to "défis technologiques". Belgium's strikes on 9 and 12 Oct are about budget cuts. KCTU's July 2026 general strike was about subcontractor bargaining, and police counted only about 5,000 at its rally.
  • Turnout is hard to verify. Even India's strikes, where organisers claim more than 250M, get cautious wire wording. AP wrote "hundreds of thousands" in July 2025 and only "partially disrupted" in Feb 2026. France reports official counts of demonstrators, not strikers. Most other countries publish no event-level striker counts.
  • The labour market has not been hit hard yet. The Economist (Sept 2026) says "the jobs apocalypse is postponed", with US unemployment at 4.1%. Still, unions increasingly frame AI as the main threat: the AFL‑CIO calls it "the single biggest threat to working people of our lifetime".

Reference class and base rate

I know of no recent strike of 1M or more in which AI or automation was a principal grievance, so the historical base rate is roughly zero. Strikes of 1M or more with verifiable turnout happen at most about once a year worldwide, mostly Indian general strikes, and even those often fail the wire-verification test. The question becomes: (chance a mass strike makes AI central) × (chance turnout is verified).

Paths to YES

  1. India. The central trade unions add AI-driven job losses as a headline demand after AI disrupts IT, back-office, banking or public-sector jobs. A national strike follows and AP or Reuters says "millions". Rough estimate: about 2%.
  2. Europe, Korea or Latin America in a fast-AI world (2029–31). A general strike explicitly against AI layoffs, or for an AI dividend or shorter hours, reaches 1M and is confirmed by a wire or official count. Rough estimate: about 2%.
  3. Other routes. The US May Day 2028 strike push; only about 830,000 US contracts expire March–July 2028. Also a sector-wide strike in a very large workforce, such as Indian banks or scheme workers, or gig platforms if "algorithmic control" is read as automation. Together about 1%.

Paths to NO

General strikes stay focused on government policy (labour codes, budgets, pensions, housing). AI demands stay as one item in long lists or limited to individual firms and sectors. Large AI-related job losses, if they come, may weaken unions' ability to strike rather than trigger strikes. And turnouts above 1M usually rest on organiser claims that the wires only pass on.

Strongest case against my estimate

AI may advance quickly. Unions in strike-heavy countries such as India, France, Italy, Korea and Argentina are already starting to treat AI as their main threat. Five years is a long window, and a single massive "general strike against AI job destruction" would settle the question. That is why the quarterly hazard rises over time instead of staying flat. Even so, the participation-verification hurdle and the strict "principal grievance" test keep my estimate low. Taking the opposite view would move me to about 8–10%, not much higher.

Series

In Python I built quarterly hazards rising from 0.05% (Q4 2026, when nothing AI-centred is scheduled) to 0.10–0.13% in 2027, about 0.2% in 2028, about 0.25% in 2029, 0.30–0.36% in 2030 and 0.38–0.44% in 2031. Cumulating them as 1−∏(1−h) gives 0.5% by end-2027, 1.2% by end-2028, 2.2% by end-2029, 3.5% by end-2030 and 5.1% by end-2031.

I read the question strictly: AI must be central in the common strike call, not one item in a long list. I would count "millions of workers struck" written in a wire's own voice as meeting the 1M threshold.

Sources

Resolution criteria

Opened 16 Sep 2026Not yet resolved

Question. By the last day of each quarter, has a strike or coordinated general strike involving at least 1 million workers in one country taken place, with the organising unions naming AI or automation as a principal grievance?

Definitions.

  • Strike: an actual collective, temporary withdrawal of labour by workers, including public-sector, informal-sector and self-employed workers. Partial-day stoppages count; strike authorisations, threatened strikes, lockouts, work-to-rule actions without a stoppage, and demonstrations without withdrawal of labour do not.
  • One strike or coordinated general strike: a single industrial dispute's strike, or stoppages called together by a union, union federation or joint organising coalition under a common strike call. A multi-day or rolling action counts if it is covered by that call. Separate rounds of action or unrelated disputes cannot be added together merely because they share a grievance.
  • At least 1 million workers: at least 1,000,000 distinct people who actually withdrew labour during the qualifying action within a single country. Count each person once, not worker-days, union membership, workers eligible or called to strike, people indirectly unable to work, or attendance at supporting rallies. Participation in different countries cannot be combined. An estimate of “one million” or “about one million” counts; “up to one million” does not. For a numerical range, its lower bound must be at least 1 million.
  • AI or automation: artificial intelligence, robotics, mechanisation, or automated software or machinery replacing or controlling work previously performed by people. Automation need not use AI. General objections to job losses, restructuring, digitalisation or surveillance do not qualify without an explicit connection to AI or automation.
  • Principal grievance: an official strike call, demands document or public statement by the organising union or joint organising coalition explicitly identifies AI or automation as a main, central or leading reason for the action, or explicitly identifies an AI/automation-related demand as a main or non-negotiable demand for ending it. It may be one of several principal grievances. An incidental mention in a long platform, commentary by individual participants, or a journalist's inference does not count. For a coalition strike, the grievance must appear in the coalition's common call or statement, not solely in an individual affiliate's separate platform.

Resolves YES if qualifying public sources establish that the action occurred, that its actual participation reached the threshold, and that its organisers made the qualifying grievance statement by the end of the action. These facts may be established by separate documents. The participation estimate must explicitly cover striking workers, rather than a combined total of strikers and other protesters.

Resolves NO at a horizon if the qualifying evidence has not been published by that horizon. Announced turnout targets, large membership rolls, annual national totals across separate strikes, and unsupported organiser turnout claims are insufficient.

Resolving source. For occurrence and participation, use an event-specific publication by the country's national statistical office or labour ministry, the International Labour Organization, or reporting by Reuters, Associated Press or Agence France-Presse. A wire report must report a participation estimate as its own assessment or cite an official statistical estimate; merely relaying an organiser's disputed or unverified claim is insufficient. Use the latest available official event-specific estimate in preference to wire estimates; absent an official estimate, use the lowest current estimate among these wires if they differ. Corrections available before qualification supersede earlier figures. For the grievance, use the organising union's or coalition's official website, strike notice or demands document, or a direct quotation of its authorised spokesperson in Reuters, AP or AFP reporting. These are the question's specific resolving sources rather than the developer-focused Public confirmation convention A2. The Resolution window and Resolution date convention A3 apply: resolution is dated to the publication that first completes the qualifying evidence. The Observation period convention applies to participation statistics; the action or participation period measured must end after the opening date. Once YES, the Resolution window's persistence rule applies.

Download all forecasts (JSON)