An anti-AI party wins 5% of a national vote

Where things stand (14 Oct 2026)

No party in any OECD country currently comes close to qualifying. I searched in English, German, Dutch, French, Finnish, Japanese, Korean and Hebrew and found no registered party whose programme names restricting AI as its main aim and that has any measurable support. Activist groups such as PauseAI and Stop AI lobby and protest, but they do not run for office.

The elections left in 2026 cannot produce a YES:

• New Zealand (7 Nov): the 17 registered parties seeking the party vote include no anti-AI party (Elections NZ, Oct 2026: https://elections.nz/media-and-news/2026/candidate-and-party-lists-released).
• Latvia (3 Oct): no anti-AI party among the winners (LSM, 4 Oct 2026).
• Israel (27 Oct): 38 lists filed; none is a notable anti-AI party (GIGA, 19 Sep 2026).
• Spain's snap election (29 Nov): polls show only the usual parties (Reuters, 5 Oct 2026).
• US House elections: a third party cannot realistically reach 5% nationally.
• Denmark (24 March) and Sweden (13 Sept) voted before the question opened.

AI has become a very prominent issue. After the summer 2026 incidents of AI agents going rogue and Jacob Coxon's resignation from Anthropic, Anthropic's CEO called for a slowdown (Time, 15 Sep 2026: https://time.com/article/2026/09/15/ai-anthropic-researcher-quits-coxon-slowdown/). A POLITICO/Public First poll found pluralities in the US, Canada, the UK, France, Spain and Germany favour pausing AI development (45–51%) (Business Insider, Sep 2026: https://www.businessinsider.com/ai-anxiety-sweeping-the-globe-cutting-across-party-lines-poll-2026-9).

But mainstream politicians are taking the issue over quickly:

• Sanders and Casar introduced a bill to ban superintelligence, with Ocasio-Cortez among the co-sponsors (PBS, Sep 2026).
• 40 MPs from several UK parties back a ban on superintelligence (Express, Sep 2026: https://www.express.co.uk/news/politics/2247984/list-40-mps-ai-killing-humans).
• Germany's Greens are calling for an AI moratorium (Hasepost, Sep 2026).
• New Zealand's Greens want a one-year pause on new AI data centres (Newsroom, 2 Oct 2026).
• Von der Leyen has leaned toward a pause.
• A study of speeches in 33 parliaments finds restriction is pushed mainly by the radical left and Greens, not by single-issue parties (arXiv 2609.02296).

This takeover leaves less room for a single-issue party.

Reference class and base rate

I used single-issue parties in OECD lower-house elections, from my own recall of standard results. Very few have reached 5% nationally, and most fall short:

• Reached 5%: Israel's pensioners' party Gil (about 5.9%, 2006), the Swiss Motorists' Party (5.1%, 1991), UKIP (12.6%, 2015).
• Just missed: AfD on the euro (4.7%, 2013), the Dutch farmers' party BBB (4.65%, 2023).
• Never close: the Dutch Party for the Animals (best 3.8%), anti-lockdown parties during COVID (Germany's dieBasis 1.4%; Austria's MFG got 6.2% only in a state election).
• Pirate Parties: passed 5% in Iceland and the Czech Republic only after broadening their platforms.

So in the last two decades, a newly salient issue has rarely, if ever, produced a party that both passed 5% nationally and would meet a strict "primary aim" test.

How I combined it

• 70%: AI stays a top-tier issue in several OECD countries through 2028–31.
• 50%, given that: some single-issue anti-AI party reaches about 2% or more in polls somewhere. There are many proportional systems with low barriers (Netherlands, Israel, Denmark, Japan, Korea).
• 20–30%: that party actually wins 5% or more in a national election before the end of 2031.
• 70%: its programme explicitly names restricting AI as its main aim, rather than a broader "humans first" mission.

That gives about 5–7%. I add roughly 1% for two other routes:

• a mainstream party declaring a superintelligence ban its "number-one priority" after an AI catastrophe;
• the low-salience world.

Total: about 7% by the end of 2031.

Timing

About 55–60 national elections fall in the window. Any new party needs time to organise, so:

• 2026: about 0 (all remaining contests checked).
• 2027 (Estonia, Finland, Switzerland, Poland, a possible UK vote in May): about 0.35%.
• 2028 (Korea, Australia, Lithuania, Luxembourg): about 1%.
• 2029–31 (UK, Germany, Austria, Belgium, Canada, Netherlands, Sweden, Japan, Israel and others): about 2% a year.

I built 21 quarterly hazards weighted to the election calendar and multiplied them through to get the cumulative series.

Strongest case against my number

Case for higher: this timeline's AI shock is unprecedented and still growing. New parties fuelled by social media have surged fast (BBB, Japan's Sanseito and Team Mirai). An AI disaster or mass layoffs could create a "stop AI" protest vote like UKIP's or the Brexit Party's in one of the many low-threshold countries. If that happens, 15–30% would be plausible.

Case for lower: a month after the peak of the scare, no party has formed. Mainstream parties are absorbing the issue, and the platform test is strict. Those points argue for 3–4%. I settled between the two.

Question. By the last day of each quarter, has a party whose own platform names restricting AI as its primary aim won at least 5% of the national vote in a general election in an OECD country?

Definitions.

• OECD country: a country that is a full OECD member on the final polling day of the qualifying election, as recorded in the OECD’s Members and partners directory or its accession announcements.
• General election: a regularly scheduled or snap nationwide election for the lower or sole chamber of the national legislature. Presidential elections, upper-house-only elections, by-elections, subnational elections, supranational elections, and referendums do not count.
• Party: a political party identified as such in the official election results. A joint electoral list or alliance counts only if it contests under a single identity with its own common platform satisfying the platform test; its votes cannot be attributed to individual constituent parties unless official results separately identify their votes.
• Own platform: an official election manifesto or formally adopted party programme published by the party, or deposited by it with an electoral authority, and publicly available before voting begins. It must remain applicable to that election. Authenticated archived copies count. Later platforms, individual candidates’ statements, interviews, and media characterisations do not establish the platform test.
• Names restricting AI as its primary aim: the platform explicitly describes imposing restrictions on AI as the party’s main, overriding, or number-one objective, or states that the party exists principally to pursue that objective. Merely including it among several priorities, placing it first in an unranked list, or devoting substantial space to it does not count. The designation must concern the party as a whole, not just its technology policy.
• Restricting AI: imposing enforceable prohibitions, pauses, caps, licensing requirements, or other mandatory limits on the development, training, deployment, or use of artificial-intelligence systems, including a specified class such as frontier AI. Voluntary safety measures, research funding, transparency alone, general opposition to technology, and data-centre restrictions without an explicit objective of limiting AI do not count.
• National vote: the party’s share of all valid votes cast nationwide in a single voting tier. Where a separate party-list vote exists, use that vote, aggregating all regional lists nationally; do not use or combine constituency votes. Otherwise aggregate votes for the party’s candidates across all constituencies, using first preferences in preferential systems and first-round votes in two-round systems. For multiple-vote candidate systems, use the electoral authority’s official national party-vote measure. The denominator includes valid votes for all contestants, including independents and parties below seat-allocation thresholds, not just votes in districts the party contested. Blank and invalid ballots are excluded. Use exact official totals where available; a rounded display of 5.0% is insufficient if the underlying share is below 5%.

Resolves YES when official final or certified election results establish a national vote share of at least 5% for a party satisfying the platform test. Winning seats or entering government is not required. A later official correction that first establishes the threshold qualifies on its publication date.

Resolves NO for a quarter if no qualifying result has been published by its end. Polls, projections, exit polls, provisional counts, regional vote shares, seat shares, and votes pooled across separate parties, elections, rounds, or voting tiers do not qualify.

Resolving source. Official final results published by the competent national electoral commission, returning authority, election-administering ministry, or official gazette; where national results are decentralised, a complete aggregation of final results from the legally competent constituent electoral authorities is acceptable. The platform test requires the party-authored documents specified above, not third-party assessments. These are narrower resolving sources than Public confirmation (A2). Apply Resolution date (A3), the Resolution window, and the Observation period: the final polling day must be after the opening date, and the qualifying official results must be published within the resolution window. A pre-opening platform may establish the party’s objectives if it remains applicable to the qualifying election. Once YES, always YES under the Resolution window convention.