Canada's Anti-AI Movement Is Powered by an AI News Aggregator

Stop the Datacentre is powered by AI.

Not metaphorically. Not in the tedious sense that somebody involved may have asked ChatGPT to rewrite a newsletter or produce a tasteful illustration of a server rack menacing a duck pond.

The campaign's public website is connected to a closed-source news intelligence system that ingests articles and social posts, interprets what they're about, assigns them an editorial category, and produces a written explanation for why they belong there.

Stop the Datacentre

Stop the Datacentre presents itself as a national network of local campaigns opposing proposed AI datacentres across Canada. The central site lists campaigns in cities including Hamilton, Toronto, Mississauga, Vancouver, Calgary, Regina, and Saint John. It promises that organizers can launch a new local campaign quickly, using a shared collection of maps, forms, election trackers, updates, and organizing tools.

The local sites all have roughly the same structure. There is an explanation of the proposed project. There are campaign updates. There is a countdown timer to the next municipal election. There is a map. There are tools for contacting politicians. There is a form for becoming an organizer. There is a continuously updated stream of articles and social posts called the Live Monitor.

The campaign's central argument is not ridiculous. Municipalities have rules for apartment buildings, warehouses, factories, roads, power lines, and almost every other large physical object a private company might attempt to place beside somebody’s home. Hyperscale datacentres do not fit neatly into many of those existing categories. A proposed facility can consume hundreds of megawatts. It can require new generation and transmission infrastructure. It can produce persistent noise. It can consume water. It can reshape provincial electricity planning and lock communities into private agreements that will outlive the politicians who signed them.

The economic case is not automatically convincing either. A facility can be physically enormous, electrically ravenous, and employ fewer permanent workers than the Costco across the road. A community is allowed to look at a warehouse demanding the output of a respectable power plant and ask what, precisely, it is getting in return.

The problem is not that Stop the Datacentre has concerns about datacentres. The problem is that the campaign demanding transparency around AI infrastructure is itself quietly operating through opaque AI infrastructure.

This is an anti-AI campaign

There is a possible defence here. Perhaps Stop the Datacentre is not anti-AI. Perhaps it opposes only specific industrial projects. Perhaps the campaign is narrowly concerned with zoning, water, electrical planning, noise, and municipal authority.

The public messaging makes this defence difficult. The organizer form asks supporters to choose their two largest concerns. The available categories include energy use, water and air pollution, lack of regulation, and:

I just hate AI

This is not a planning concern. It is not "I oppose hyperscale development in residential areas." It is not "I want a provincial environmental assessment." It is not "I am concerned about electricity prices." It's a generalized position toward the technology itself.

The public accounts surrounding the campaign are similarly unambiguous. They use names like No AI Vancouver and No AI Greed. Organizers have publicly called for bans on AI use in British Columbia schools. The campaign is attached to a broader political movement that does not merely want stricter zoning conditions for datacentres. It wants artificial intelligence constrained, rejected, or prohibited across substantial parts of public life.

That's all fine. People are allowed to oppose AI. There are very good reasons to oppose it, or want it regulated. But they should probably disclose when AI is operating the machinery of their own movement.

Powered by Wardroom

At the time of writing, the bottom of each Stop the Datacentre site contains a short declaration:

Stop the Datacentre 2026 • powered by Wardroom

The sentence beneath it describes Wardroom as a civic-monitoring platform developed in Hamilton, Ontario to track local issues, discussions, and municipal decision-making.

This sounds explanatory until you visit Wardroom. Wardroom is not a general-purpose civic movement platform. It's someone's closed-souce intelligence dashboard for Hamilton municipal politics. It monitors a specific city. It follows local news, political discussion, council activity, and public information concerning Hamilton. It is presented as a finished application, not as a reusable framework.

There is no public repository or installation guide. There is no documentation describing how to create a new tenant, define a political campaign, deploy a municipal microsite, manage supporters, publish election trackers, or generate maps of proposed industrial developments. It is not a general-purpose content management system or civic organizing framework.

Saying that Stop the Datacentre is "built on Wardroom" therefore makes very little technical sense on its face. You can build a campaign site with Drupal. You can build one on NationBuilder. You can build one using a general-purpose application framework or a hosted organizing platform. You do not normally build a national political campaign on top of a closed municipal intelligence dashboard for Hamilton.

Not unless the public dashboard is only one interface into a much larger private system. Either "powered by Wardroom" is being used so loosely that it means little more than "made by the same people," or Wardroom is not the underlying platform at all. It is one deployment of a broader ingestion, classification, publishing, and intelligence system that is not publicly documented.

The Stop the Datacentre sites strongly suggest the second explanation. They do much more than display Hamilton news. They operate across multiple cities. They collect supporter information. They publish local campaign material. They map development proposals. They track elections. They expose shared RSS and JSON feeds. They combine manually selected stories with automatically classified news and social posts.

That is not a municipal politics dashboard being repurposed, there's a platform here. The public Wardroom site simply is not the platform being described. Something sits underneath it.

"Other Projects"

Wardroom has a small "other projects" link tucked away in the corner of the page. It goes directly to a site called Newscube.

Newscube's homepage is sparse in the particular way early-stage technology companies believe conveys seriousness:

Systems for live editorial intelligence

The site explains Newscube develops hardware and software for tracking and interpreting information environment. It lists two products. The first is "Radar":

Live news intelligence surface focused on movement, continuity, and situational awareness.

The second is Newscube itself:

A dedicated hardware interface for live editorial intelligence and one-click content distribution across social channels.

This is startup language. There's a product category, a software product, a proposed hardware interface. There is a carefully constructed phrase, "live editorial intelligence", describing the thing being sold.

Radar appears to remain in a closed preview. There is no working surface, pricing page, public documentation, or visible path for someone outside the project to begin using it. The startup page states, "Radar is slowly opening up to the public. If you already have an account, sign in. If you have a code, create one." and requires an invite code to sign up.

That does not prove Newscube is incorporated, venture-backed, or currently charging customers. "Startup" here describes the visible shape of the operation: a closed-source product family, a named market category, a preview-stage flagship application, and a live deployment demonstrating what the technology can do.

That live deployment appears to be Stop the Datacentre.

What Newscube sells

Newscube's copy does not describe a typical RSS reader or news aggregator. Newscube says it tracks and interprets public information environments. Interpretation is the product. The system is meant to take a large stream of news, social posts, political discussion, and public information, then reduce it into something operationally useful.

That is precisely what Wardroom does for Hamilton municipal politics. It watches the city, gathers local material, and turns that material into a clean intelligence dashboard. Stop the Datacentre then performs the same operation for a national political campaign. The campaign’s Live Monitor gathers stories and social posts about datacentres. Its feed divides material into pinned, social, and news sources. Some entries are selected manually. Others are processed by an automated classifier.

The machine beneath the monitor

Each Stop the Datacentre site includes a Live Monitor. The visible interface looks simple enough. It is a stream of news stories and social posts related to datacentres. There are headlines, excerpts, images, publication names, and timestamps. Below the monitor are links to RSS and JSON feeds.

Naturally, I opened the JSON.

The top level describes a "datacentre bucket", lists that the sources are three type of material: ["pins", "social", "news"], and reports seperate totals for each.

This already tells us more than the public interface. The site is not merely displaying a hand-maintained reading list of content about data centres. It has distinct systems for manually pinned items, social posts, and news stories.

The manually selected records are easy to identify. Their IDs begin with pin:, they have a pin_id field, and include "pinned": true.

Then there are the other records. One contains a Bluesky post by Hamilton journalist Joey Coleman discussing legal advice released by Hamilton City Council concerning an AI datacentre. The record includes the original post, its publication time, its source feed, the time it entered the system, an excerpt, and four additional fields:

"bucket": "datacentre", "rationale": "The post discusses legal advice related to an AI data centre and mentions municipal election coverage.", "classifier": "agent", "classified_at": "2026-07-17T02:25:03+00:00"

The post was published to Bluesky at 02:14 UTC. The feed reports it as arriving at 02:15. The agent classified it at 02:25. The timestamps imply an asynchronous job. The post was published, ingested by the system, and queued for a background classification job which finished 10 minutes later. The system took an unstructured social-media post, interpreted its meaning, assigned it to the datacentre category, and generated a plain-language explanation for the decision.

That is an automated semantic classification pipeline. It is AI in every reasonable sense of the word.

This is not a keyword filter

It would be possible to build a datacentre monitor without AI. A conventional program could subscribe to RSS feeds and search for words such as datacentre, data centre, hyperscale, megawatt, GPU, cooling, or the names of known developers. It would be extremely crude, but it would also work reasonably well.

The public feed shows the system doing more than that. A keyword filter can determine that a post contains a particular word. It does not ordinarily produce a sentence explaining that the post discusses legal advice related to an AI datacentre and municipal election coverage.

The system is making a contextual judgment. It is interpreting unstructured language, assigning an editorial category, and creating new metadata explaining why the category applies.

The feed is careful not to identify a specific model, instead just using the generic term "agent". It does not name OpenAI, Anthropic, Google, Mistral, or any other provider. It does not exp

[truncated for AI cost control]