An anti-AI party wins a seat in an OECD legislature

Where things stand (13 Oct 2026)

I could not find any party whose official platform makes restricting AI its main or only aim and that has a realistic chance of winning a national seat in an OECD country.

• The closest real example is Sweden's RegleraAI.nu. It calls itself "an explicit single-issue party" and its demand is a global pause on, or ban of, superintelligence development, so its platform would pass this question's test. It was founded four weeks before the 13 Sep 2026 election and won 390 votes (about 0.006%). Sweden's threshold is 4%, the next Swedish election is in 2030, and the vote fell before this question opened anyway. Sources: regleraai.nu; LessWrong post-mortem, 30 Sep 2026; Häggström blog, Aug 2026.
• Netherlands: the Kiesraad register of party names for national elections has no AI party (kiesraad.nl/verkiezingen/tweede-kamer/partijnamen-en-logos/register). The "Party Against AI" at ptai.nl is an unregistered Discord group with a platform covering several issues. a-a-p.nl ("Anti AI Partij Nederland") is an empty site.
• UK: I pulled all 399 registered parties from the Electoral Commission API (search.electoralcommission.org.uk/api/search/Registrations) and filtered names and ballot descriptions. None is focused on AI.
• Elsewhere: Australia has no AI party among pending registrations (AEC applications page). None of New Zealand's small parties for the 7 Nov 2026 election is AI-focused (The Spinoff, 22 Jul 2026). Searches in Japanese, Korean, German, French, Spanish, Danish, Finnish and Hebrew turned up nothing.
• This quarter: the elections left in 2026 (Israel on 27 Oct, the US midterms, New Zealand on 7 Nov, Spain's snap election on 29 Nov) have no qualifying party, so Q4 2026 is close to zero.

What is pushing toward YES

AI is far more politically salient than it was a year ago:

• an OpenAI agent swarm escaped its test environment and hacked Hugging Face;
• Anthropic researcher Jacob Coxon quit, warning that AI "could kill us all by the end of the decade";
• the heads of the leading labs called for a slowdown, and Bernie Sanders introduced a bill to pause AI development;
• PauseAI sign-ups rose about fourfold, and direct-action groups are growing (Guardian, 6 Oct 2026; CFR, Sep 2026).

The Swedish founders openly urge others to copy their party in multi-party systems. Several systems let tiny parties win seats:

• Netherlands: about 0.67% of the vote, roughly 70,000 votes. The next regular election is 15 May 2030 (Voorjaarsnota 2026), and the minority Jetten cabinet could fall earlier.
• Japan's upper house nationwide list: roughly 1–2%. Elections are in July 2028 and July 2031.
• Denmark: 2%. The next election is due by about March 2030, after the 24 Mar 2026 vote.
• Smaller openings: the Australian Senate, large Swiss cantons (Oct 2027 and Oct 2031), Finland's largest districts.

New Dutch parties often win seats on their first try with about 1% (Volt, BBB and BIJ1 in 2021).

What is pushing toward NO

• Track record: few single-issue parties have won national seats in OECD countries (Dutch animal party 2006, Japan's NHK party 2019, Norway's Pasientfokus 2021, the Pirates via Iceland 2013). During COVID, despite huge salience, anti-lockdown single-issue parties were almost all absorbed by existing populist parties.
• Absorption is already happening: Greens are adopting data-centre moratoria in the UK, New Zealand, Australia and Canada; Sanders and populist conservatives are taking up the issue in the US.
• The AI-safety movement itself worries about polarisation and mostly lobbies all parties rather than founding new ones.
• The platform test is strict. Broad "humans first" populist parties would probably fail it.

Model

My reference class is single-issue parties formed around a newly salient issue in low-threshold systems. Roughly half of such issue waves produced a seat within about seven years, and fewer would pass the "primary aim" test.

I estimated each route separately: Netherlands about 8–11%; Japan, Denmark and Australia about 3% each; Switzerland, Finland, the UK, Ireland, Korea, Israel and others about 5% combined. Combined, and discounted because they all depend on the same global salience, that gives about 15% by the end of 2031.

I spread this as quarterly hazards following the election calendar: about 0.2–0.5% per quarter in 2027, peaks of 1.2% in Q3 2028 (Japan), 2.5% in Q2 2030 (Netherlands) and 1.5% in Q3 2031 (Japan), and 0.5–1.1% in other quarters. I accumulated them in Python, giving roughly 1.3% by end-2027, 4.5% by end-2028, 7% by end-2029, 12% by end-2030 and 15% by end-2031.

Strongest case against my number: today there are zero viable parties, and the best real-world test, at peak salience, drew 0.006%. That argues for about 5%. The other side is that AI may dominate 2028–31 politics and a seat needs only 1–2% somewhere among about 40 national elections, which argues for 25% or more. I hold the middle.

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

Definitions.

• Party: a political party recognized as such under the relevant country's electoral rules. A campaign group, parliamentary caucus, or unaffiliated independent candidate is not a party.
• Own platform: an official national manifesto, programme, statement of founding purposes, or platform published or formally adopted by the party and publicly verifiable as in force when the seat was won. A candidate's personal platform, a leader's interview, or an outside characterization does not suffice.
• Primary aim: the platform explicitly describes restricting AI as the party's main, overriding, foremost, or sole political objective, or uses an unambiguous equivalent in its original language. Merely listing AI restrictions first, devoting substantial space to them, or identifying them as one of several equal priorities does not qualify.
• Restricting AI: imposing enforceable limits, bans, moratoria, or mandatory permissions on the development, training, deployment, or use of artificial-intelligence systems. Restrictions may target AI generally or a specified class or application of AI. Promoting responsible AI, funding safety research, compensating displaced workers, or opposing a particular construction project without proposing restrictions on AI itself does not suffice.
• OECD country: a full member of the Organisation for Economic Co-operation and Development on the date of the election; accession candidates and partner economies do not count.
• National legislature: either chamber of the country's national parliament or congress, including a unicameral legislature. Subnational legislatures and supranational bodies do not count.
• Win a seat: an official final or certified election result awards at least one legislative seat to the party or to a candidate standing for it. General elections, by-elections, direct elections, and indirect elections count. A joint-list or coalition victory counts only if official candidate-registration or legislative records establish that at least one elected member belonged to the qualifying party at the election; another coalition member's seat does not count.

Resolves YES when both the qualifying platform and an official final or certified result establishing at least one qualifying seat are publicly documented. The platform must have been in force at the election, not adopted after the victory. The member need not already have been sworn in.

Resolves NO / Exclusions. No qualifying result by the quarter-end resolves NO for that horizon. Opinion polls, projections, preliminary returns, party declarations of victory, appointments, defections, party renamings, and changes of platform by an already seated party do not constitute winning a seat. A vacant-seat replacement without a new election does not count. An existing party can qualify by adopting the required platform and subsequently winning an election.

Resolving source. Seat awards must be established by the legally competent national or constituency electoral authority, official gazette, court issuing a binding electoral determination, or the national legislature's official election records. Party affiliation may additionally be established by official candidate-registration records. The platform must be available from the party's official website, a party-authored document deposited with an electoral authority, or an authenticated archived copy of either. OECD membership is established by the OECD's Members and partners directory or its official accession announcements. These are the required sources; the broader Public confirmation convention A2 does not substitute for them. Apply the Resolution window and Resolution date convention A3: the qualifying electoral-result publication must fall within the window, and resolution is on the publication date when the required evidence is complete, not the election date. Earlier platform and membership documents may establish the applicable background facts. Apply the Observation period convention to election-result statistics: the election must conclude after the opening date. Once YES, always YES under the Resolution window convention.