Investing/Nov 15 edition/Reporting & analysis
RegulationStocks

RegulationRules, enforcement & market structure

Voters Back Data Center Restrictions in 46 of 47 Local Ballot Measures as Retailers Block AI Shopping Agents

A weekend roundup headlined by Dror Poleg's essay on AI eroding privacy also points to firmer developments: voters backed restrictions in 46 of 47 local data center measures this year, and major retailers are blocking AI shopping agents to protect add-on revenue.

From the Abnormal Returns article of November 14, 2026 and cited sources: Ballotpedia counts 46 of 47 data center ballot measures in 2026 as favoring restrictions. [3] [5]

SOURCES[3] Voters have decided on 47 data center measures —… news.ballotpedia.org[5] Voters rebuff data centers in every ballot measure except… cnbc.com

AI-assisted, not human-edited. Methodology Corrections Who owns this site

KEY FACTS
What happened
Ballotpedia counts 46 of 47 data center ballot measures in 2026 as favoring restrictions. [3] [5]
Original source
Abnormal Returns article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
Studies
arXiv 2507.18253 (Jul 2025) [8] date from arXiv ID
When
November 14, 2026 article published
Evidence
11 sources across 11 sites

Reach: not measured yet

THE CORE IDEAS4 TAKEAWAYS
01

Ballotpedia counts 46 of 47 data center ballot measures in 2026 as favoring restrictions. In the November midterms, voters ruled against data centers in 40 of 41 measures across 10 states, 21 of them in Ohio. The only exception was Harwood, North Dakota. More votes come in December, including a 180-day moratorium vote in Hollis, Maine, and a vote on banning large data centers in Mandan, North Dakota. [3] [5]

02

Coverage disagrees on what the votes mean for the AI infrastructure buildout. One New York Times headline linked in the roundup says data centers will still get built. Yahoo Finance asks whether the roughly $1 trillion boom is in trouble. [1] [11]

03

Amazon has blocked AI shopping agents, including Meta's Muse and OpenAI's ChatGPT bots. Target and The RealReal have set their own, different limits. An agent can return an incomplete product comparison without telling the user, so developers are advised to test against specific retailers instead of assuming access. Tomasz Tunguz argues that merchants who depend on add-ons and fees gain nothing from bots that skip the extras. He cites United, whose upgrade and fee revenue exceeds its operating income. Shopify, by contrast, takes a cut of each transaction and welcomes agents. [4] [7] [6]

04

The roundup's privacy theme comes from Poleg's essay. It argues that strong AI will push people to act as if everything they do could eventually be exposed, and to hide their real thoughts more carefully. The essay is philosophical and cites no empirical evidence. Academic work such as 'Countering Privacy Nihilism' argues against giving up on privacy. [1] [2] [8]

WHY IT MATTERS

The ballot results and retailer blocks are documented, but their effects are not. Agent builders may hit site-by-site access gaps that open-web benchmarks miss.

Read the full assessment

Data center developers may face more local resistance, though coverage disagrees on whether the buildout will slow.

Executive brief

The privacy headline on this roundup comes from an essay with no data in it. The surrounding AI links are where the numbers are. Abnormal Returns' Saturday roundup quotes Dror Poleg saying that with strong AI, people should act as if everything they do can eventually be exposed. Poleg's essay is philosophical and cites no studies. The same roundup links to more concrete AI stories: retailers blocking AI shopping agents to protect their upsell revenue, and voters rejecting data centers in 46 of 47 local ballot measures in 2026, per Ballotpedia.

What changed and event timeline

  1. Retailers block AI shopping agents

    Amazon blocked Meta's Muse agent, and Target and The RealReal set their own, different limits. Agents can return incomplete product comparisons without the user knowing,, which cites The Verge and Mashable.

  2. Data centers lose at the ballot box

    Voters went against data centers in 40 of 41 midterm measures across 10 states. Ohio had 21 of them. The one exception was Harwood, North Dakota ().

  3. Poleg's "Under the Eye of AI."

    Poleg argues that AI's ability to hack systems and erode privacy will make people hide their real thoughts more carefully, not become more honest ().

  4. Tunguz on why merchants block bots

    Tomasz Tunguz argues that merchants whose profits depend on add-ons and fees, such as Amazon and United Airlines, gain nothing from automated buyers that skip the extras ().

  5. Abnormal Returns roundup

    The weekly roundup leads with Poleg's quote and also links Tunguz and a New York Times story saying data centers will still get built despite the election ().

Capabilities and access

None of the reviewed sources names a model or version.

Read the full section

None of the reviewed sources names a model or version. The concrete access fights are:

  • Amazon keeping AI agents off its sites, including Meta's Muse (Let's Data Science) and OpenAI's ChatGPT bots (eMarketer headline).
  • Shopify welcoming agents, because its roughly 3.1% cut of each transaction pays out whoever the buyer is (Tunguz).

Technical analysis for researchers and developers

  • Agent reliability depends on the site. Benchmarks run on open sites can overstate how well an agent works in practice.
  • Blocks are often invisible to users. Agent designs need to detect blocks and say so.
  • Bot traffic data is fragmented. Tunguz combines figures from Imperva, HUMAN Security, Cloudflare Radar and a16z, and these vendors measure traffic in different ways.
Read the full section
  • Agent reliability depends on the site. Developers are advised to treat retailer access as a condition that varies by site, not a built-in capability, and to test agents against specific retailers before launch (Let's Data Science). Benchmarks run on open sites can overstate how well an agent works in practice.
  • Blocks are often invisible to users. If a block isn't reported, the agent's answer can look complete when it isn't. Agent designs need to detect blocks and say so.
  • Bot traffic data is fragmented. Tunguz combines figures from Imperva, HUMAN Security, Cloudflare Radar and a16z, and these vendors measure traffic in different ways.

Claims and evidence

  • Bots are 54% of global web traffic in 2026, and AI agent traffic rose 8x from August to October 2026, according to Tunguz citing vendor reports.
  • Tunguz says United's $4.8B in seat and bag upgrades plus $3.2B in fees is more than its $4.7B operating income.
  • 46 of 47 data center measures this year favored restrictions (Ballotpedia).
Read the full section
  • Bot share of traffic. Bots are 54% of global web traffic in 2026, and AI agent traffic rose 8x from August to October 2026, according to Tunguz citing vendor reports. For comparison, Imperva's 2025 report put bots at 51% (ITPro). Both figures come from security vendors, not independent measurement.
  • Merchant profit dependence. Tunguz says United's $4.8B in seat and bag upgrades plus $3.2B in fees is more than its $4.7B operating income. He also says Amazon's $68.6B in ad revenue produces profit roughly equal to its retail operating income (Tunguz). He cites 10-K filings, but the ad-profit split is his own calculation; Amazon does not report ad profit separately.
  • Ballot results. 46 of 47 data center measures this year favored restrictions (Ballotpedia).
  • Privacy. Poleg's thesis is an argument without empirical support in the essay (Poleg).

Context and prior work

  • Academic work on privacy and AI includes Countering Privacy Nihilism, which argues against giving up on privacy, a counterweight to Poleg's framing.
  • Forrester has warned that blocking every bot can also block shoppers acting for real customers (Forrester).
  • Shopify has set limits on "buy for me" bots on merchant sites, so the tension predates 2026.

Limitations, safety and contested findings

  • Poleg's essay is theological and philosophical.
  • The reviewed coverage disagrees on what the votes mean. The NYT headline in the roundup says centers will still be built.
  • CNBC reports that 56% of voters called data centers a crucial or major factor in their vote (CNBC). Exit-poll wording varies by pollster.
Read the full section
  • Poleg's essay is theological and philosophical. It draws on Deuteronomy 18:13 and a Hasidic commentary, and gives no evidence about how AI actually exposes people (Poleg).
  • Data center backlash vs. buildout. The reviewed coverage disagrees on what the votes mean. The NYT headline in the roundup says centers will still be built. Yahoo Finance asks whether the roughly $1 trillion boom is in trouble.
  • Exit polls. CNBC reports that 56% of voters called data centers a crucial or major factor in their vote (CNBC). Exit-poll wording varies by pollster.
  • Corroboration. No other outlet in the reviewed sources covers the roundup itself.

Business and practitioner implications

  • Agent builders: expect access to vary retailer by retailer, and be open with users when a site has blocked the agent.
  • Merchants: Tunguz suggests rethinking pricing or charging a checkout fee for agents rather than blocking them (Tunguz).
  • Infrastructure planners: more local votes are scheduled for December, including a 180-day moratorium vote in Hollis, Maine on Dec 8 and a vote in Mandan, North Dakota on banning large data centers on Dec 15 (Ballotpedia).
Read the full section
  • Agent builders: expect access to vary retailer by retailer, and be open with users when a site has blocked the agent.
  • Merchants: Tunguz suggests rethinking pricing or charging a checkout fee for agents rather than blocking them (Tunguz).
  • Infrastructure planners: more local votes are scheduled for December, including a 180-day moratorium vote in Hollis, Maine on Dec 8 and a vote in Mandan, North Dakota on banning large data centers on Dec 15 (Ballotpedia). In Ohio, Governor-elect Amy Acton backs conditions such as requiring data centers to pay 100% of their utility costs (CNBC).
FOLLOW THE EVIDENCE

The source trail.

Sources (11)
A LITTLE LESS NOISE. A LOT MORE CONTEXT.

Stay curious.
Follow the evidence.

Independent perspectives, the original sources, and room for the questions that don't have easy answers.

How we build the brief
Connect with us

Find us where you already read.