Quick Take · News · Put AI to Work · Looking Ahead

Quick Take

Oracle just cut up to 30,000 workers. They did it by email, at 6am, with no prior warning from managers or HR. Access to company systems was cut immediately. The company posted $6 billion in quarterly profit. The reason for the layoffs: Oracle has committed $156 billion to building AI data centers and needs to free up cash.

Not because AI replaced those workers' jobs. Because Oracle is betting it might, eventually, and wants the money now. (The Harvard Business Review has a term for this: layoffs based on AI's potential, not its actual performance.)

In response, Sanders and Ocasio-Cortez introduced a bill to ban all new data center construction until Congress passes AI safety, worker protection, and environmental legislation. Over 100 communities have already enacted local moratoriums on data center construction. The frustration is real.

But so is the pushback. Sen. Mark Warner, at an Axios summit the same day: "A data center moratorium simply means China is going to move quicker." Sen. Maria Cantwell called for renewable energy investment, not a construction freeze. Even Democrats who support stronger AI regulation aren't backing this bill.

Both sides are describing real problems while proposing incomplete answers. A moratorium doesn't stop the AI race. It relocates it. And unregulated construction doesn't create broad prosperity. It creates Oracle-style layoffs to fund buildings that employ 50 people while communities absorb the infrastructure costs.

Progressives have always had the answer to this kind of fight. We didn't ban factories. We demanded safety standards, living wages, and environmental rules. We didn't ban cars. We required seatbelts and emission controls. The AI infrastructure boom needs the same treatment.

Build it right, or don't build it at all.

Let's get into it.

---

AI News Roundup

The data center fight: moratorium, myth, and the middle ground

What happened: The Sanders-AOC Artificial Intelligence Data Center Moratorium Act (S. 4214) would freeze all new data center construction until Congress passes separate legislation covering AI safety reviews, worker protections, environmental limits, and civil rights. Sanders: "We cannot sit back and allow a handful of billionaire Big Tech oligarchs to make decisions that will reshape our economy, our democracy and the future of humanity."

The bill reflects real frustration. Data center electricity costs have jumped as much as 267% in some areas over the past five years. Communities bear infrastructure costs while tech companies collect tax breaks. Oracle just showed what the current arrangement looks like for workers: 30,000 jobs cut at a profitable company, no advance warning, all to fund construction.

But the moratorium has virtually no path to passage, and most Democrats oppose it. The competitiveness argument is straightforward: China commissioned 30+ hyperscale data centers in 2026 alone. The EU is fast-tracking AI infrastructure permitting. The Middle East is spending hundreds of billions on sovereign compute capacity. A U.S. pause doesn't slow global AI development. It just moves the construction overseas.

The temperature myth, and why honest data matters: You may have seen the headlines last week. "Data Centers Are Creating Heat Islands." "AI Data Centers Raise Temperatures Up to 6 Miles Away." CNN, Fortune, and Gizmodo all ran versions of a Cambridge University preprint claiming data centers raise local temperatures by up to 9.1 degrees Celsius and affect 340 million people.

The study was not peer-reviewed, and the methodology has a basic flaw. Researchers measured land surface temperature via satellite, which captures how hot a rooftop or parking lot gets in direct sunlight. By that measure, any large building looks like a heat source. A Walmart, a warehouse, an Amazon fulfillment center. (It turns out buildings are hotter than grass. Who knew.) Experts at The Register pointed out the study had no control group comparing data centers to other commercial buildings. Andy Masley ran the actual physics: data center waste heat accounts for roughly 1 to 3 percent of the observed warming signal. Everything else is the generic urban heat island effect that happens when you replace grass with concrete and steel.

Remember the water panic? Same pattern. Data centers use about 3.3% of what U.S. golf courses consume nationally. Google's entire global data center operation uses less water than 43 golf courses. Arizona alone has 287, and nearly half of them exceed their legal water limits every year with no penalty.

None of that means data centers get a pass. The environmental concerns are real. Electricity costs near data centers have surged. Water use matters in drought-prone regions. Communities deserve transparency about what's being built and what it costs them.

But exaggerated claims do real damage. They give industry lobbyists an easy target. "See? The critics don't even have their facts straight." Honest proportions make the real concerns harder to dismiss. And the real concerns are serious enough on their own.

The answer progressives should be championing: Community benefit agreements with teeth.

Brookings published a framework in January laying out what these should include: data center operators contributing at least 0.5% of gross revenue to a locally administered community fund, local hiring requirements at living wages with apprenticeship programs, closed-loop cooling and water consumption caps, public dashboards tracking jobs and resource use, and mandatory infrastructure upgrade funding.

This is already happening at the state level. Illinois introduced the POWER Act, which would require data centers to fund their own renewable energy generation and disclose water usage. Colorado has competing bills from labor and environmental groups, but both sides agree on the principle: build it right, with good union jobs and clean energy. They're arguing about enforcement mechanism, not values. That's a good argument to be having.

The NAACP launched a "Stop Dirty Data Centers" campaign with an organizing playbook for frontline communities. They've partnered with local organizations in Prince George's County, Maryland to challenge a task force that rubber-stamped data center expansion without examining environmental justice impacts. Their framing: data centers are a civil rights issue when they're disproportionately sited in communities of color and low-income neighborhoods. (See our Put AI to Work section below for how to use these resources to research what's being proposed in your area.)

---

Deepfake political ads are now standard campaign strategy

What happened: In the 2026 midterm cycle, AI-generated attack ads have moved from isolated incidents to routine tactics. The NRSC released an 85-second deepfake of Texas Democrat James Talarico that appeared to show him adding mocking commentary to his own tweets. The "AI GENERATED" disclosure appeared in small text for roughly three seconds. In Georgia, Rep. Mike Collins released a deepfake of Sen. Jon Ossoff saying things he never actually said. In Texas, Sen. John Cornyn ran a deepfake music video attacking primary rival Ken Paxton and posted another deepfake of Rep. Wesley Hunt with no AI disclosure at all.

There is no federal law prohibiting any of this. Twenty-eight states have passed deepfake disclosure laws, but campaigns are treating the disclosure requirement as a legal fig leaf. The tiny text, the three-second flash, the label nobody reads — technically compliant, functionally useless. Researchers studying the 2026 cycle found that roughly 50 percent of voters said deepfakes influenced their decisions even when they knew the content was AI-generated.

Why this matters: This is not a hypothetical threat that might affect future elections. It is happening in real races right now, and the people making these ads face no consequences worth worrying about. The deepfakes land emotionally before anyone thinks to question them. By the time a viewer notices the fine-print label, they've already formed an opinion.

---

In brief

The EU is rolling back its AI Act before it even takes effect. The European Commission's "Digital Omnibus" package would let AI companies self-certify their risk assessments instead of publishing them, rewrite GDPR to make it harder to remove personal data from AI training sets, and create "disproportionate effort" loopholes for data removal requests. Amnesty International called it "an unprecedented rollback of rights online." The Corporate Europe Observatory documented how Big Tech lobbying shaped the changes article by article. This matters for U.S. advocates because the EU AI Act was the strongest evidence that real AI regulation is achievable. If it gets gutted before the August 2026 enforcement deadline, it weakens that argument everywhere.

The ICE surveillance system from Issue 4 just got ruled unconstitutional. In Issue 4 (March 3), we covered Palantir's ELITE targeting app and ImmigrationOS, the system using Medicaid records to find deportation targets. ICE officers have now confirmed under oath in federal court that ELITE identified their arrest targets and that they operated under quotas of 8 arrests per team per day. A federal judge ruled the warrantless arrest pattern in Oregon unconstitutional under the 4th and 5th Amendments. The system we documented is now the subject of active federal litigation.

---

---

Put AI to Work

Practical ways progressives can use AI this week

Find out what's being built in your backyard

The data center boom is a local fight. Zoning changes, tax incentive packages, water permits, and environmental impact assessments all happen at the county and municipal level. You don't need to be a policy expert to engage — you just need information, and AI can help you get it fast.

Step 1: Find the proposals. Go to your county's planning commission or zoning board website. Search meeting agendas and public notices for "data center," "colocation facility," or "enterprise computing." Many counties publish agendas as PDFs. Download them and upload to Claude or ChatGPT. Ask: "Are there any data center or large-scale computing facility proposals in these documents? What stage are they at? What's being requested?"

Step 2: Analyze the deal. Most data center proposals come with tax incentive packages. Find the economic development agreement or tax abatement application (usually a public record, sometimes requires a records request). Upload it and ask: "What tax breaks are being offered? What does the community get in return? Are there local hiring requirements? Renewable energy commitments? Water usage limits? How does this compare to the Brookings community benefit agreement framework?" That last question gives you a benchmark to measure the proposal against.

Step 3: Check the environmental picture. Search your state's environmental agency database for any permits related to the site. Ask AI to help you understand them: "This is a water withdrawal permit application for a data center in [county]. How much water are they requesting? How does that compare to current usage in this area? What are the cooling system specifications?" If the NAACP data center playbook covers your region, cross-reference their findings.

Step 4: Show up prepared. Before a planning commission meeting, compile what you've found and ask AI to help you draft three-minute public comment remarks. Give it your findings, the Brookings CBA benchmarks, and the Illinois POWER Act requirements as comparison points. Ask for specific, constructive recommendations rather than blanket opposition. Planning commissioners respond to residents who bring alternatives, not just objections.

Step 5: Share your research. Write up your findings as a one-page community brief. Share it with local environmental groups, your city council member, and local news outlets. A well-researched summary of what a data center will actually cost your community — with honest numbers — is more powerful than a petition with 500 signatures.

The NAACP playbook, the Brookings CBA framework, and the AI Now Institute's data center policy toolkit are all free and public. Thirty minutes of AI-assisted research and you can walk into a planning meeting better prepared than most developers expect you to be.

---

---

Looking Ahead

One thing keeps showing up in the stories we cover: The technology is not the variable, the rules are.

AI scribes help burned-out doctors (Issue 7). AI surveillance helps ICE target families through their Medicaid records (Issue 4, now ruled unconstitutional). AI deepfakes are poisoning elections with no legal consequence. AI data centers are creating real economic opportunity and real environmental costs. Same technology, every time. Different outcomes depending on whether anyone wrote the rules.

The progressive position on AI infrastructure should be the same as on every other industry: Build it right. Community benefit agreements that ensure local hiring at living wages. Renewable energy requirements that don't shift costs to residential ratepayers. Water transparency that lets communities make informed decisions. Enforceable standards, not voluntary commitments.

We've been making this argument about factories, power plants, and housing developments for a century. It works. Three state legislatures just proved it works for AI too, with specific bills passing on bipartisan or unanimous votes.

The playbooks are public, the organizing tools are free, and your planning commission meets next month. Show up.

Until next time,
Jordan

---

Read past issues on the web · Subscribe via RSS · Website