AI Factions: A Political Field Guide

A mapping of six of the major factions shaping AI and the political landscape: their beliefs, money, policy positions, schisms, and alignments.

Kate Gage, kate@higherground.institute

1

The Six AI Factions

Kate Gage, kate@higherground.institute

2

The Sixth Faction: AI Populists

Every other faction is defined by its answer to "what is AI?" This one is defined by "who has power?" Its analysis: most people are open to AI as technology — but not to the tech oligarchy, corruption, autocracy, and climate cost that come with its current trajectory. That makes it a political-economy faction, not an AI-prediction faction.

  • Theory of change: near-term harms (surveillance, deepfakes, job loss) are legitimate concerns AND organizing gateways to structural demands about power and wealth concentration.
  • The data center moratorium movement is its main people-power lever: slowing the buildout opens the Overton window for a real policy framework. (Not universally embraced — some in the camp think it is the wrong fight to attach to AI regulation.)
  • A "token tax" is the emerging insider version of the same strategy: policy pressure to bring labs to the negotiating table.
  • Expect many more groups to land here: this is the accountability wing of the progressive movement arriving in AI politics.

Kate Gage, kate@higherground.institute

3

Terminology Decoder

Language as a tell for different factions (a sampling- many more shibboliths exist)

Kate Gage, kate@higherground.institute

4

The Political Overlay: Money, Coalitions and Flashpoints

The Money (2026 cycle)

AI industry PACs and nonprofits have raised more than $185M this cycle. Only $55.3M has been spent on candidates, which means most of the money has not landed yet. Transformer AI in Politics Tracker

  • Pro-industry: Meta's network is the largest at $110M+ (American Technology Excellence Project, Forge the Future, Making Our Tomorrow, plus a California PAC). Leading the Future has $75.8M disclosed. (Donors: Marc Andreessen and Ben Horowitz at $25M each, Anna and Greg Brockman at $12.5M each, Perplexity at $100K - four people and one company accounting for 99% of the raise.) Build American AI, an LTF-linked 501(c)(4) with no disclosed donors, holds roughly $50M and launched a new electoral arm, Building the Future, in August 2026. Innovation Council Action pledged $100M that has not materialized.
  • Pro-guardrails: Public First Action ($20M from Anthropic, its only disclosed donor), led by former Reps. Brad Carson (D-OK) and Chris Stewart (R-UT), and its three PACs plus AnthroPAC. Plus You Can Push Back, Dream NYC, Guardrails Alliance ($742K) (largest single donor is an OpenAI employee at $200K) and Encode AI (~$120K). Non-industry money is under 2% of the total.
  • First stress test, NY-12 Democratic primary: $29.5M flowed around RAISE Act author Alex Bores in an eight-candidate field that also included Micah Lasher, Jack Schlossberg and George Conway. Pro-safety PACs spent $21.3M for Bores (Jobs and Democracy $15.0M, You Can Push Back $3.4M, Dream NYC $2.5M, Guardrails Alliance $395K). Think Big, a Leading the Future affiliate, spent $8.2M against him. Bloomberg separately put roughly $10M into a super PAC for Lasher, his former aide, with no stated AI rationale. Lasher won 39% to 35%. The safety side outspent the industry side 2.6 to 1 and still lost, which makes NY-12 a better story about the limits of AI money generally than about industry money specifically.


Flashpoint 1: Federal Preemption of State AI Laws

The Tech Right wants one national framework and tried a state-law moratorium and NDAA preemption. A horseshoe coalition beat it: Warren and Sanders alongside Bannon, Greene, Hawley, and Paul. The July 2025 amendment stripping the moratorium from the reconciliation bill passed 99-1, sponsored by Sens. Cantwell, Blackburn, Markey, and Collins; 290+ state lawmakers had already opposed preemption. A separate NDAA preemption attempt followed — these are distinct fights. A December 2025 executive order then created an AI Litigation Task Force to sue states anyway.


Flashpoint 2: Data Centers and Electricity Prices

Bipartisan backlash has made data centers a kitchen-table issue: three-quarters of Democrats and nearly two-thirds of Republicans oppose a nearby facility, and utility bills are now a core affordability concern. Dozens of projects have stalled or been blocked (see figures below).

Bottom line: safety, ethics, and AI populist camps hold no power in this White House; the Tech Right has money and access but keeps colliding with a populist-progressive horseshoe over state preemption, data centers, and now surveillance.

$129.7M

AI PAC money raised but not yet spent on any candidate [Transformer, Sept 2026]

99-1

Senate vote on the Cantwell-Blackburn-Markey-Collins amendment stripping the state-law moratorium from the reconciliation bill

61%

Voters oppose a data center near where they live, with 14% in support [Annenberg, June-July 2026, n=1,320]

75 projects

$130B in data center projects blocked or delayed in Q1 2026 alone [Data Center Watch]

Kate Gage, kate@higherground.institute

5

Schisms and Surprising Overlaps

Internal Fights Worth Knowing

Within AI Safety

Yudkowsky/MIRI purists ('shut it all down') vs. Coefficient Giving/Anthropic . Anthropic is the flashpoint: hypocrite to some, best hope to others.

Within the Right

Tech Right vs. MAGA base is open war over preemption and now surveillance. The Tech Right's own voters are turning against data centers and rate hikes.

Within the Left

Ethics camp and safety camp compete for regulatory oxygen; the TESCREAL fight leaves open wounds. The AI Populists' data-center moratorium strategy is not universally embraced even inside the camp.

Within Labor

Building trades want data center construction jobs and resist moratorium framing. AFT and AAUP are funding the Guardrails Alliance. SAG-AFTRA and WGA are fighting a separate consent-and-compensation war over training data. AFL-CIO has not unified these blocs into a single AI position.

Within the Child Safety Coalition

Agreement on need for accountability; disagreement on remedies. Fairplay, Encode, and Design It For Us want duty-of-care and age-assurance mandates. EFF, CDT, and Fight for the Future call that an ID checkpoint on the internet and oppose KOSA and STOP CSAM outright. CDT co-signed an anti-age-verification letter with Americans for Prosperity.

Kate Gage, kate@higherground.institute

6

Scrambler 1: Data Centers

Data centers are where the factions break formation. And the fight is no longer two-sided: alongside the camps that want to stop the buildout and the camps that want to accelerate it, a third group is writing the rules for how it happens. Who pays, what gets taxed, what gets disclosed, what gets enforced.

🛑 STOP

Ethics and enviro left — 500+ groups from 47 states demand a congressional moratorium.

"Wherever these hyperscale data centers are built, electricity rates skyrocket for families and small businesses, water stability is threatened, and communities are turned upside down." — Emily Wurth, Food and Water Watch

AI Populists — the moratorium is strategy, not just defense: slowing the buildout is the one people-power lever that opens space for a real regulatory framework.

"Data centers increase residential energy bills by upwards of 250% and many communities are left in the dark on these projects from the start." — Meghan Pazik, Public Citizen

MAGA populists — blasting their own administration:

"How dare the peasants complain about data centers stealing their water and driving the cost of electricity!!" — Marjorie Taylor Greene, mocking reported FBI monitoring of data center protesters

Sanders and Ocasio-Cortez introduced the AI Data Center Moratorium Act; roughly a dozen states have introduced moratorium bills; New York's executive order is the first statewide pause; Maine's LD 307 passed both chambers and was vetoed; Monterey Park, CA became the first US city to permanently ban data centers by ballot measure.

AI Now's North Star Data Center Policy Toolkit is written for this posture, with scaffolded fallback protections where a full stop is not feasible.

🏗️ BUILD

Tech Right / White House — The AI Action Plan's core pillar is "Build American AI Infrastructure": faster permitting, looser environmental review.

"Technology is going to happen. Trying to stop it is like ordering the tides to stop." — David Sacks, White House AI czar

Frontier labs, including the "safety" one — OpenAI wants 250 GW of capacity by 2033, "a gigawatt of new AI infrastructure every week." Anthropic committed $50B to US data centers, asking "do we have what it takes to build AI in America?" (though it pledged to shoulder ratepayer price hikes itself).

The White House Ratepayer Protection Pledge (March 2026) is the industry's answer to the affordability backlash, but it is voluntary and unenforceable, with no interconnection, financing, or supply chain mechanism behind it.

There is no organized constituency for building. Polling runs roughly 30 for and 70 against in both parties, which is what happens when one side stops making a case (Jasmine Sun).

🧰 HOW

The largest and least organized group: not pro-build, not pro-moratorium, looking for a framework. Most of the 2026 policy activity lives here, with 300+ data center bills filed across 30 states in the first six weeks of the year.

Who pays. Strict cost causation, special rate classes for large loads (18+ states), take-or-pay commitments, and decommissioning bonds.

Tax. Virginia enacted a first-of-its-kind $0.011/kWh data center electricity consumption tax. Others are pausing incentives or conditioning abatements on grid connection and clean power. Federal insider variant: the token tax.

Environment and enforcement. Emissions get decided in rate cases, air permit hearings, and enforcement, not in model design: prohibit new on-site gas, require clean firm power, mandate load flexibility, register large loads with NERC. Climate Power scores the eight hyperscalers on behind-the-meter gas and local grid carbon.

Process and consent. The backlash is driven less by water and noise than by process failure and weak benefits: NDAs, fast tracks, and being blindsided (Jasmine Sun). So: ban NDAs binding public officials (Microsoft has already conceded this), require pre-permit disclosure, and set standard terms of entry. New York's draft floats $1M in community benefits per megawatt.

The grid-leverage case: treating the buildout as yes/no is a strategic error, and there is a narrow, closing window to convert data center demand into transmission, ratepayer relief, and domestic supply chains (Jane Flegal, Searchlight).

The governance case: the moratorium push is a strategic dead end with real equity costs, and the answer is democratic oversight of how AI develops (Holly Buck, Jacobin).

Kate Gage, kate@higherground.institute

7

Scrambler 2: Open Source AI

"Open weights" is not open source: you can download the model but not the training data or code. Critics call the gap "open washing." Each faction means something different by "open."

Open as Weapon

Tech Right: a16z leads with "Asserting American Leadership in Open Source AI"; it is in Trump's AI Action Plan. The pitch: open models let little tech compete and beat China. The tell: "open source" here mostly means deregulation.

Open as Threat

AI Safety: you cannot recall an open model; guardrails can be fine-tuned away. Geoffrey Hinton compared open-sourcing big models to "being able to buy nuclear weapons at Radio Shack." This drove the SB 1047 fight (Newsom veto, Sept 2024).

Open as Commons

Public interest left: Mozilla is spending from $1.4B in reserves on an open source "rebel alliance" against OpenAI and Anthropic ("owners, not renters"), with the Public AI movement pushing publicly funded models and compute. But AI Now warns "open" models still run on Big Tech's compute.

The China Effect

DeepSeek (Jan 2025) made open models a geopolitical weapon. Chinese open models passed American ones in downloads (1.15B vs 723M by March 2026). The US response: OpenAI's gpt-oss, the ATOM Project — while Meta went proprietary ("the Great Re-Closing").

Why It Matters

Openness splits the allies (safety vs ethics camps); every regulation fight now contains an open source carve-out battle; the Tech Right borrows the left's "democratize AI" language. Read any open source claim with three questions:

  1. Open what — weights, data, code, or brand?
  1. Who profits?
  1. What regulation is this argument being used against?

Kate Gage, kate@higherground.institute

8

Scrambler 3: Surveillance and Flock Cameras

The Setup

Flock Safety runs 100,000+ AI license plate cameras across the US, an $8.4B company whose lead investor is a16z, which led Flock's $275M Series F in March 2025 at a $7.5B valuation via its American Dynamism fund. Flock integrates with Palantir, which holds a $30M ICE contract. Marc Andreessen joined the Pentagon's Defense Policy Board in June 2026. Nearly 800 cities signed new Flock contracts in 2026.

The Blowup

Local camera data flowed to ICE and federal agencies without authorization. Dayton's audit found its cameras searched 7,100+ times for immigration enforcement, explicitly against city policy. An Illinois state audit found CBP had been given access to Illinois plate data through an undisclosed pilot program, forcing dozens of agencies off the state's data-sharing system. Ring ended its partnership with Flock on February 12, 2026, following public backlash after its Super Bowl ad. At least 53 cities across 20 states had canceled or rejected Flock contracts as of June 2026, with independent trackers putting the figure at 80 or more by mid-2026. Cities include Austin, Evanston, and Eugene.

Opposition from the Left

"It doesn't matter which company has its creepy cameras in your neighborhood, they all have the same problems: a lack of transparency, oversight, and regulation into how they collect, store, and use our data." — ACLU, whose national campaign is literally called "Get the Flock Out"

The EFF calls Flock's Nova product, which links plate data to personal information, a "surveillance panopticon."

Opposition from the Right

Rep. Lauren Boebert introduced the Surveillance Accountability Act restricting warrantless tracking; a bipartisan amendment would condition federal highway funds on banning license plate readers except for tolls; Newsweek now runs lists of Republicans cracking down on Flock.

The Schism

The Tech Right's money built the surveillance network; the MAGA base and progressives are fighting it city by city. For AI Populists this is the cleanest "who is in charge" story in AI politics: VC money → police cameras → ICE via Palantir.

Kate Gage, kate@higherground.institute

9

Methodology, Caveats and Open Questions

How This Map Was Drawn

Factions are defined by shared beliefs, community infrastructure (funders, conferences, forums), and political behavior (money, endorsements, coalitions), building on Michel Justen's faction taxonomy, with original additions: the political money overlay, the AI Populists faction (from field feedback), and the issue-line analysis.

Caveats

  • Snapshot as of September 2026. PAC totals, alignments, and flashpoints move fast.
  • Factions are clusters, not cages. Individuals vary; orgs may dispute their placement.
  • Placements reflect public positions and funding, not private views.
  • Some fights are coalitional, not factional. On child safety and surveillance, the table shows positions, not alliances.

How the orgs we've placed would place themselves

We invite corrections. Placements reflect public positions and funding; private views may differ and this is not intended to suggest relationships between groups listed alongside each other.

What we're watching for/need to update for this guide:

National Security/International Cooperation

Not simply US/China competition, but also opportunities for international cooperation around safety, other international efforts

Labor placement

Where does labor belong, and how to differentiate? Some trades are pro-data-center jobs; some are funding safety efforts; there isn't a central alignment

Child safety as the preemption vehicle

Watching whether the carve-out survives a real federal bill: KOSA, the CHATBOT Act, and the Youth AI Privacy Act cleared Senate Commerce on August 5, 2026, and the open question is what preemption language rides along, and whether the age-verification split splinters the anti-preemption coalition.

Tech Right surveillance collision

Tracking the Tech Right's surveillance investments that could collide with its own voter base as the MAGA right is increasingly opposed to warrantless tracking.

Kate Gage, kate@higherground.institute

10

Sources and Further Reading

Primary Explainers

Interactive Maps

  • Mapping AI (mapping-ai.org/map) — ~1,800 people and orgs in US AI policy by stance and funding
  • CSET / MIT AI Risk Initiative — governance landscape taxonomy

Political Money and Flashpoints

  • NYT: "New Super PAC, the Guardrails Alliance, Aims to Rally Tech Workers to Help Limit AI" (June 2026)

Kate Gage, kate@higherground.institute

11