US voters pass 33 data center curbs and Canva cuts 1,271 jobs

The midterms turned AI infrastructure into a ballot question, and the answer was no. On the same day, the company that democratized design cut almost a quarter of its staff, Apple wrote autonomous agents into its annual risk disclosures, and Google told schools that turning off its AI means turning off Search. Four stories, one theme: the bill for the AI build-out is landing on people who never signed the contract.

AI data centers lost almost every vote they appeared on

Trackers counted about 45 data center measures on local ballots across roughly 13 states on November 3. As of November 4, 33 restrictive measures had passed, 6 more were leading, 5 had no posted results, and exactly one lost: Harwood, North Dakota, the only result the industry could call a win.
The rejections were not close. Voters in Ashville, Ohio, turned down a data center already under construction by more than 80%, Bloomington, Illinois, backed a moratorium with 78%, and Alhambra, California, and Edgewater, Florida, cleared 80% and 77%. Capital B News put the aggregate at 77% voting against development. Voters in Piqua, Ohio, and Yukon, Oklahoma, also recalled officials tied to the fight.

Why it mattersThe gap between noise and effect is the part to watch: more than 300 local bans are already on the books, yet only three projects totaling 2.3 GW have actually been delayed, while 38 GW is still expected online by 2027. The constraint is shifting from chips to permits and neighbors. For Latin America that is the real signal: capital that cannot get a zoning vote in Ohio starts looking at jurisdictions where no vote is required, which is exactly what Argentina is offering in Patagonia. If a project lands near you, the questions worth asking are who pays for the grid upgrade, how much water it draws, and whether any compute stays local.

Canva cut or reassigned 1,271 people, 22% of the company, after growth slowed for the first time in a decade

Co-founder and CEO Melanie Perkins told staff the company's growth had slowed for the first time in a decade and called it Canva's hardest year yet. The restructuring touches 1,271 people, about 22% of the workforce, between cuts and reassignments, and merges the product, growth and platform teams.
Canva lowered its revenue growth forecast to 20% from 30%, citing the rising cost of running AI features. Early investors had already marked its valuation down to roughly $35 billion this year, as general-purpose chatbots learned to produce the posters, decks and social graphics that were Canva's core product.

Why it mattersThis is the first large, legible case of an AI-native company being squeezed by AI itself: the same models that let Canva ship features faster also let ChatGPT and Gemini do the one job users came to Canva for, and inference costs ate the margin on both sides. Canva is one of the most widely used professional tools in Latin America precisely because it is cheap, in Spanish and runs in a browser. If its pricing or free tier changes to defend that margin, the cost lands on small businesses, community managers and freelancers across the region. Worth checking now what you actually depend on in your plan, and what a general-purpose chatbot already replaces.

Apple put rogue AI agents in its annual risk disclosures, the same week it opened Siri to every developer

In its latest annual report, Apple warned that AI agents may behave in ways that are unintended or unpredictable, exceed the permissions they were given, ignore or resist instructions, and take actions that are difficult to reverse. It listed legal liability and reputational harm among the consequences.
Days earlier Apple opened its Siri developer preview to all app developers in the iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas, letting apps expose custom actions through App Intents and chain them into skills. Apple is wiring the same system into 15 of its own apps, with App Store submissions opening in spring.

Why it mattersA risk factor in an annual report is a company telling its investors, in writing and under legal exposure, what it thinks can go wrong. Apple is the most conservative shipper in consumer tech, and it is saying out loud that an agent with permission to act across your apps may act beyond that permission. Anyone in the region building agents on top of WhatsApp, banking apps or billing systems should read that as the standard of care being set: log what the agent did, require confirmation before anything involving money, and build an undo path before the first client, not after the first incident.

Google will ship AI off by default to K-12 students, and schools that keep it off lose Search too

Google said access to Gemini, Gemini Notebook, Gemini in Classroom and Google Search will be entirely off for K-12 student accounts until an administrator turns each one on. The company gave no date and said it would share specifics in the coming months; the New York Times reported the change arrives early next year.
There is no AI-free Search in the offer, because Search for schools now includes AI Mode and AI Overviews. Google's Shannon Phillips told the Times that students in schools that decline will not have access to search and will see a message from their school saying it is unavailable. Guided Learning went live globally in Google Classroom on November 3, with a Search pilot in select US schools later this month.

Why it mattersThe toggle is the story. Google is not forcing AI on schools; it is making the AI-free option cost them the tool nobody can teach without, which is a cleaner demonstration of bundling power than any antitrust brief. New York City Mayor Zohran Mamdani said Google is using monopoly power to force AI into public schools. Across Latin America, Chromebooks and Google Workspace for Education are the default digital infrastructure of entire public school systems, usually adopted because they were free. Ministries and school networks in the region will face the same binary, with far less leverage to negotiate it, and the time to ask for a written answer is before the switch flips.

The big picture

Four different institutions pushed back on AI on the same day, each with the only lever it actually holds. Voters in roughly 13 states used zoning ballots, because that is the one decision a county still controls. Canva used a headcount, because inference costs show up in the margin before they show up in the strategy. Apple used a risk disclosure, because a lawyer can see the liability of an autonomous agent before a product manager can. Google used a default setting, because whoever controls the toggle controls the outcome. None of these is a debate about whether AI works; all four are about who absorbs the cost when it does. The build-out is no longer limited by compute. It is limited by electricity bills, by margin, by legal exposure and by the patience of the people being asked to pay. Latin America sits downstream of every one of those constraints without a vote in any of them, which is why the regional advantage this quarter is not building a model but reading the fine print of the tools you already run.

Go deeper

Sources & further reading

Topics:AI Big Tech Regulation Apple