The Digital Divide The Digital Divide

← Newsletter archive

Issue No. 7 · Week of October 5–11, 2026

Someone Finally Said the Quiet Part

By Rowa Taha

This week, the actual numbers came out from under the press releases — a president's own arithmetic on what one data center costs his country, a count of how few AI Act enforcement actions exist a month in, and two courts telling two companies exactly what they can't make people accept.

Data Centers — United States

AI Data Centers Are Tearing Towns Apart — Literally

Local opposition to AI data centers has curdled into open threats: a CNN review of more than 100 police reports found escalating threats against data centers nationwide, from bomb claims to online calls to "use bullets on city council members," reported Oct. 12. The same week, a coalition of 185 community, labor, and climate groups urged Congress to reject a "sweetheart deal" for data centers they say is buried in permitting-reform legislation. A UMass Amherst poll found just 11% of Americans would support an AI data center in their own community, versus 65% opposed — this isn't fringe opposition anymore, it's the median position, and it's the predictable result of siting AI infrastructure with a fraction of the community input a highway or power plant would require.

CNN → Center for Biological Diversity → Stateline →
Data Centers — United Kingdom

Betting Big on Data Centers While Scotland Votes to Block Them

The UK government wants to triple the country's data-center capacity by 2030 — from roughly 2GW today to as much as 6.3GW — even as a power crunch and grid-connection queues threaten to delay the buildout, and projected water use could hit 42 million litres a day, twice the capacity of a reservoir currently under construction to serve it. Scotland's parliament didn't wait to find out: it voted 80–26 on Sept. 16 to block planning decisions on all 24 proposed hyperscale data centers in the country until the Scottish Government publishes national guidance, after campaigners noted their combined power draw at full build-out could exceed Scotland's entire winter peak demand. Westminster and Holyrood pulling in opposite directions on the same infrastructure says plainly that "necessary for AI" and "good for the community hosting it" are not the same claim.

Data Center Dynamics → Estates Gazette → Friends of the Earth Scotland →
Accountability — South Korea / Japan

AI Just Lowered the Bar for Hacking Two Countries at Once

A wave of AI-assisted cyberattacks hit named companies across two countries this month — SoftBank, Daiwa Securities, and convenience-store chain Lawson in Japan, plus nine banks and two megachurches in South Korea — with South Korean ransomware incidents up 76.8% in the first half of the year alone. AI doesn't just lower costs for defenders; it lowers the skill floor for attackers, and neither country's disclosure regime was built for that math.

NBC News → Insurance Journal → Beinsure →
AI Governance — India

New Delhi Promises an AI Rulebook, No Law Yet

India's IT Minister Ashwini Vaishnaw said this week the government will release a public consultation paper on AI regulation — covering safety, deepfakes, and a stated "human-first" framing — within a month. That's a promise of a draft, not a law, and India has a documented habit of "coming soon" AI governance news that takes far longer than a month to actually arrive.

Organiser → Careerindia → The420.in →
Accountability — European Union

The AI Act Turns One Month Old. GDPR Is Still Doing the Actual Enforcing.

As of Oct. 2, independent trackers count just 14 enforcement actions and 3 open inquiries across the EU that reference an AI system at all — and every fine issued so far was built on GDPR's existing powers, not the AI Act's own, since the Digital Omnibus on AI pushed the Act's high-risk-system obligations back from August 2026 to December 2027. The "world's most comprehensive AI framework" is, one month into its headline enforcement date, still mostly a framework; the law doing the actual work predates it by eight years.

AI Enforcement Tracker (vorplabs) → AI Act Enforcement Tracker → EU Regulatory Enforcement Tracker →
Data Rights — Nigeria

A Nigerian Court Said Meta Can't Make Tracking the Price of Using Instagram. Meta Is Appealing.

In a 46-page judgment delivered Sept. 25, Justice A.F. Pokanu of the Lagos High Court ruled that Meta's non-consensual profiling of Nigerian Facebook and Instagram users for behavioral advertising violated their constitutional right to privacy under Section 37 of the 1999 Constitution and the Nigeria Data Protection Act. The court ordered Meta to stop the practice, bring its Nigerian operations into compliance within eight weeks, and pay the applicants the naira equivalent of $100,000 in damages plus costs. Meta filed its appeal five days later, teeing up a bigger question for Nigeria's Court of Appeal than the damages themselves: whether a platform can legally require users to accept behavioral tracking as the price of access at all, rather than offering it as an actual choice.

Businessday NG → TheNigeriaLawyer → TechCabal →
Data Centers — Kenya

A President Said the Quiet Part Out Loud: "We'd Have to Shut Down Power for Half the Country"

Microsoft and Abu Dhabi's G42 have been negotiating a $1 billion, 1-gigawatt AI data center in Olkaria since 2024 — a single facility that would draw roughly a third of Kenya's entire 3,000-megawatt installed capacity. Talks stalled earlier this year over annual lease-payment guarantees Kenya couldn't commit to, and they remain unresolved as of this week's renewed regional coverage. President William Ruto's own framing is the bluntest admission yet, from any head of state, of what hosting this infrastructure actually costs: "To switch on that one data center, we would need to shut down power for half the country."

Yahoo Finance / Reuters → Data Center Dynamics → Techweez →
Lawsuit — United States / Argentina

Four More Plaintiffs Just Joined the AI-Copyright Pile-On

The Seattle Times and Newsday sued OpenAI and Microsoft on Sept. 4 over training data scraped from behind their paywalls; WikiHow filed its own suit Aug. 22; and Argentina's Editorial Perfil, publisher of Perfil and the Buenos Aires Times, became one of the first non-U.S. outlets to sue OpenAI directly. An Argentine publisher suing an American AI company says something the U.S.-only framing of most AI-copyright coverage misses: this isn't a domestic publishing-industry dispute, it's every country whose press feeds these models and sees none of the return.

Engadget → The Spokesman-Review → Unite.ai →
Lawsuit — United States

Google's Ad-Tech Punishment, Finally Public

Follow-up to Issue No. 3: the sealed remedies in the Google ad-tech case were unsealed Sept. 16, and the actual terms have more teeth than the "no breakup" headline suggested: real-time bid-data sharing, Prebid interoperability requirements, a bar on Google's own ad network bidding directly into its ad-server auction, and a six-year independent monitor with full access to Google's staff, systems, and source code, effective globally within 60 days.

AdExchanger → PPC Land → PYMNTS →
AI Governance — Council of Europe

The World's First Binding AI Treaty Still Only Has One Real Member

More than a year after opening for signature, the Council of Europe's Framework Convention on AI — the first legally binding international AI treaty, open to EU and non-EU countries alike — has exactly one party that's actually ratified it: the EU itself, in May. Switzerland signed back in March 2025 but still hasn't drafted the implementing legislation it needs, with a self-set deadline of "by the end of 2026"; the treaty needs five ratifications, including three Council of Europe states, before it takes effect for anyone.

Council of Europe (official) → CMS Law → Regulations.AI →
— End of Issue No. 7 —
Back to the archive →