A candidate pledged to follow an AI's recommendations wins office

Where things stand (14 Oct 2026)

I found no candidate anywhere who has made a qualifying pledge, has a realistic chance of winning, and is running in an eligible race whose result will be published in the rest of 2026.

• Victor Miller ("VIC"/"Victor"), Wyoming. He promised to hand 100% of his decisions to an AI as a "meat avatar". He was kept off the 2026 US Senate ballot, and a federal judge dismissed his lawsuit on 29 Sep 2026. In his 2024 run for Cheyenne mayor he got about 300 votes. His nephew is running for Cheyenne city council, which is not an eligible office. (Cowboy State Daily, 13 Oct 2026; Cowboy State Daily, 4 Jun 2026; Wyoming Tribune Eagle, 21 May 2026)
• Gaitana IA, Colombia, March 2026. It got fewer than 3,000 votes (under 2%) for an Indigenous seat. It was a tool for tallying community consensus, which the criteria exclude. (Latin America Reports, 17 Mar 2026)
• Path to Rebirth, Japan. This party has an "AI leader", but it handles party operations, not lawmakers' votes. The party won 0 seats in 2025 with 524,787 proportional votes, about half of what one seat needs. Its former leader lost the Komae mayoral race in June 2026 by 4,683 votes to 19,603, and members have left over the AI-leader idea. (CNN, 19 Sep 2025; J-CAST, 23 Mar 2026; Yomiuri, 28 Jun 2026)
• Denmark's "AI Partiet – Leder Lars". Its party name is approved until 8 Sep 2028 (Danish Interior Ministry list, 1 Oct 2026). A search snippet from the signature site showed 4 voter declarations against about 20,182 needed. The page did not load, so I could not confirm that count.
• Team Mirai, Japan. It won 11 lower-house seats in February 2026, but it uses AI as a tool and humans decide, so it does not qualify. Steve Endacott, who ran as "AI Steve", now says AI should never make political decisions (SmarterUK).
• Remaining 2026 races. Toronto's "AI mayor" candidate, Weizhen Tang, only proposes AI for budget transparency and is a fringe candidate (The Local). I found nothing relevant in Ottawa, the US midterms, Taiwan, the Brazil runoff, or Spain's 29 November snap election (Reuters, 5 Oct 2026).

Base rate

From 2017 to 2026 there were about 10 "AI politician" efforts: SAM, Alisa, Matsuda in Tama, the Synthetic Party, AI Steve, VIC, Markun and Lex, Gaitana, Victor, and Path to Rebirth. None won. Results were tiny: 179 votes for AI Steve, about 300 for VIC, 627 for Markun (Época Negócios, 7 Oct 2024), and under 3,000 for Gaitana. Several never got on the ballot. Only about one to three per year ran for an eligible office, and each got under 1–2%. Public opinion is also against the idea: fewer than 10% of Californians support government agencies using AI to make decisions that affect them (Carnegie, 17 Sep 2026). OpenAI and Meta have cut off accounts used by AI candidates.

How it could resolve YES or NO

YES most plausibly comes from a gimmick or protest list that wins a seat in a low-threshold system. Examples: the Netherlands (about 0.67% for a seat), Japan's upper-house national list (about 2%, elections in 2028 and 2031), Brazil's open lists (2030), South Africa (2029), or Philippine party-list seats. A celebrity or anti-corruption populist adopting an "AI decides" pledge is another route. Fast AI progress before 2031 could make such pledges more appealing. NO is the default. The pledge test is strict: a named system, followed as the general rule, with no free veto. Most tech-friendly politicians hedge with "AI advises, humans decide", and backlash against AI is growing.

Forecast

I put the chance at about 1.5% if AI progress stays ordinary (65% weight), about 8% if it is fast (25%), and about 15% if it is transformative (10%). That gives roughly 4.5% by the end of 2031. I spread this as quarterly hazards and computed the cumulative series in Python. The hazard starts at about 0.05% for the rest of 2026, when no viable candidate exists. It rises to about 0.3% per quarter in 2030–31 and is weighted toward big election quarters, giving a cumulative 4.4%.

Strongest case against my estimate: there are thousands of elections and some very low thresholds, so with fast AI progress my number could be too low. A second risk is a lenient reading of borderline "AI-synthesised constituent input" pledges. On the other side, the record so far is zero wins and vote shares far too small, which could justify about 2%.

Question. By the last day of each quarter, has a candidate who pledges to decide according to the recommendations of a named AI system won election as a head of government, national legislator, governor, or mayor of a city of more than 1 million people?

Definitions.

• Pledge: an explicit, public commitment made before the decisive election vote, in the candidate's published platform or other candidate-authorized campaign material, to follow the named AI system's recommendations as the general rule for substantive official decisions or legislative votes. A commitment covering all or most such decisions qualifies; a commitment limited to a single issue, pilot, or occasional consultation does not. Exceptions for legal obligations, technical failure, or emergencies are permitted; unrestricted discretion to reject recommendations on policy grounds is not. The pledge must not have been publicly withdrawn before the decisive vote. Actual compliance after election is not required.
• Named AI system: an identifiable AI model, product, or custom agent designated by name in the pledge, from any developer. A named custom agent built on an existing model qualifies without specifying its model version. References only to 'AI,' algorithms, evidence, or expert advice do not qualify. AI used only to transmit or tally human instructions does not qualify.
• Eligible offices: national head of government (including a president who heads the executive government); voting member of either chamber of a national legislature; governor or equivalent executive head of a first-order subnational jurisdiction, such as a state, province, or prefecture; or mayor or equivalent head of a municipal government. The population threshold applies only to mayors. Ceremonial heads of state, party leaders, deputy offices, local councillors, and regional legislators do not qualify in those capacities.
• Win election: be officially declared elected to the eligible office by the competent electoral authority, including through a party-list seat allocation, an electoral college, or an election by a legislature or municipal council. An appointment without an electoral vote does not qualify. A nomination, primary victory, runoff qualification, or party victory without an individual seat or office award does not qualify. A human candidate standing as the representative of an AI system qualifies.
• City of more than 1 million people: the municipal jurisdiction governed by the elected mayor has strictly more than 1,000,000 usual residents according to the latest official census or resident-population estimate published before the decisive vote. Use the national statistical office's figure, or an official municipal or regional statistical authority's figure if none is available nationally. Use municipal boundaries, not the larger metropolitan area, unless that entire area is the mayor's legal jurisdiction.

Resolves YES when both the qualifying pre-election pledge and an official final declaration or certification of that candidate's election are publicly documented. The election result must be published within the Resolution window; the underlying campaign pledge and population baseline may predate it. A later breach of the pledge, resignation, or reversal of the election result does not undo a YES under the Resolution window convention.

Resolves NO if no such case has been documented by the quarter-end. Merely promising AI-assisted government, promoting AI development, adopting AI-written campaign material, or promising to consider recommendations does not qualify. Losing campaigns do not qualify: for example, Brighton & Hove City Council's 2024 Brighton Pavilion result records Steve AI losing, not winning.

Resolving source. The competent national, regional, or municipal election commission, returning officer, electoral court, or official legislative/council record must establish the election result. Candidate- or campaign-authored platforms, manifestos, written pledges, or official recordings/transcripts establish the pledge; archived copies are acceptable. Official statistical publications establish population. These are the required sources rather than the general Public confirmation (A2) source menu; press descriptions alone are insufficient. Apply Resolution date (A3): resolve on publication of the last required qualifying record, not polling day or inauguration day. The Observation period convention does not apply to the historical population baseline used to classify the jurisdiction.