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

Quick Take

Most professionals use AI the same way they use a vending machine. Type a prompt, take the answer, leave. If the answer is mediocre, they shrug and write the email themselves.

The pros getting real value out of AI work it like a thinking partner. They iterate. They tell it the specific context of their work. They push back when it's vague. They ask it to critique its own answer before they accept it.

This month, three teachers in the AFT showed exactly what that looks like: Gabriela Aguirre, a first-grade dual-language teacher in San Antonio, used ChatGPT, Gemini, and Claude to build bilingual English/Spanish flashcards for her students. AFT teachers in New York are building agentic tools to monitor IEP compliance. None of them took the first answer. All of them taught the AI their context. That's the difference.

Let's get into it.

---

Progressive AI Win — We'll take it.

OpenAI killed Sora — their text-to-video app, and a major source of AI-generated slop.

The consumer app shut down on April 26. The API closes September 24. For a tool that was supposed to define the AI video moment, the wind-down is fast and quiet.

Be honest about why: it wasn't democracy. Sora was burning roughly $1 million a day. Its user base had cratered to under 500,000. Sam Altman is redirecting the compute to compete with Claude Code. OpenAI didn't kill Sora to protect the 2026 election. They killed it because it was bleeding money.

But the result is real, and we'll take it.

NewsGuard tested Sora 2 against twenty known false claims. It generated convincing video for sixteen of them, including five Russian disinformation narratives. The watermark could be stripped in four minutes with a free tool. Public Citizen called the system out by name as a 2026 midterms threat back in November.

We already have three Republican deepfake ads in the cycle: Talarico in Texas, Ossoff in Georgia, Spanberger in Virginia. There's no federal law on this! (Twenty-eight states have something, but most of them only require disclosure.)

So one less easy-to-use political-deepfake tool walking into the next six months is a real win, even if it's an accidental one. The caveat: the tools that survived — Runway, Veo, Hedra — don't have meaningful election guardrails either. Sora is gone. The field still has no floor.

Take the win. Then keep pushing for regulation.

---

More wins this week

WGA's new contract landed Tuesday. The Writers Guild ratified its 2026 MBA on April 24 with 90.4% voting yes. On top of the AI guardrails from 2023, the new contract adds notification and compensation requirements when studios use writers' work to train AI systems. Big union, fresh win, clean precedent.

Washington passed the strongest worker AI law in the country. HB 1672 requires human oversight for employment decisions made with AI, bans emotion-prediction AI in the workplace, and carries $10,000 penalties per violation. Companion bills (HB 1770, HB 2225) added AI content disclosure requirements and chatbot safety rules.

Illinois HB 3773 is now in force. Employers using AI in hiring decisions have to give applicants notice. It's a precedent more states can build on.

Three wins, three weeks — advocacy works.

---

Put AI to Work and stop renting SaaS by the seat

I'm not a developer, but I've still built more than fifty tools with Claude Code over the last few months. Almost every one of them replaced something I used to pay a SaaS company for.

Anthropic just made this kind of work easier: Claude Cowork lives on the desktop, reads and edits your local files, and runs multi-step tasks across your apps. Available on every paid plan. Anthropic's own pitch is the right frame:

Most AI tools are built around the prompt. Claude Cowork is built around the outcome.

So the move is two-step.

Step 1. Set up Cowork (or Claude Code or ChatGPT Codex if you want to go even further). Then ask it how to set itself up to help your specific work. Tell it about your work and ask it to find ten things it could automate for you. Iterate on the one that surprised you.

Step 2. Once it's running, point it at one of these. Each one is a SaaS bill I'm not paying anymore.

You're paying for...	Roughly	What I built instead
Social media scheduler Buffer, Hootsuite, Sprout	$30–100/mo	An N8N + Claude pipeline that posts to Bluesky and Twitter, exactly the cadence I want. How I built it →
Task manager Asana, Todoist, Things	$5–15/mo per seat	Drift — built in one session, ADHD-friendly, lets me bounce between projects without losing track. /ai → Drift
Domain monitor DomainTools, RegistrarSafe	$10–200+/mo	Domain Locker — tracks expirations, DNS, SSL, and WHOIS for the 47 domains I manage across personal and client work. /ai → Domain Locker
Email/newsletter platform Mailchimp, Constant Contact	$30–300/mo	Sendy + custom templates running an online community after AWS SES denied production access. /ai → Email Infrastructure

The full list of fifty-something projects is at jordankrueger.com/ai. The point isn't to copy any one of them. It's that you're allowed to build the tool you actually need instead of paying someone else's monthly fee for 80% of what you wanted.

---

---

Looking Ahead

• May 13 — Colorado AI Act enforcement deadline. The country's only law requiring algorithmic discrimination accountability is under DOJ attack. Worth watching.
• September 24 — Sora's API officially closes. One less surface for political deepfakes. The other tools still need work.

If you start treating AI like a coworker instead of a vending machine, your output changes pretty quickly. Try it this week. Tell it what you actually do. Ask it what it could take off your plate. Iterate on the one that surprised you.

Until next time,
Jordan

---

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

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

Quick take

Last week at ActionKit ClientCon, I sat on a panel about how progressive organizations should approach AI. Someone in the audience asked whether we should be using these tools at all, given the ethical concerns: data scraping, labor displacement, environmental cost, corporate consolidation. The concerns are real. I said:

I don't think the people we're fighting for will thank us in 20 years for standing the moral high ground and not using the best tools available to us.

Here's what I've been thinking about since: the moral high ground is already a fiction inside our own movement.

78% of progressive nonprofits used generative AI in their fundraising, marketing, or advocacy last year. Only 42% have a policy. That's 216 progressive orgs in the M+R Benchmarks data, and we see similar governance gaps in other nonprofit sector surveys too.

In org after org, the practical debate about whether to use AI is already over. Staff are using Claude, ChatGPT, Otter, Canva AI, whatever their work requires, sometimes on personal accounts, often without clear guidance, because clear guidance doesn't exist. The gap isn't adoption. It's permission. The costs of that gap are real, and they're growing.

This issue is about closing it. Let's get into it.

---

AI news roundup

The app launch boom is a permission-gap story in disguise

What happened: App Store and Google Play releases jumped 60% year-over-year in Q1 2026, and the pace has accelerated since. April is up 104% worldwide, with iOS alone up 89%. Apple's marketing chief Greg Joswiak summed it up: "Rumors of the App Store's death in the AI age may have been greatly exaggerated." TechCrunch points to AI coding tools like Claude Code and Replit as the main driver, and the builders leading the surge aren't professional developers. They're people with ideas and a problem to solve, writing software for the first time because they finally can.

Why this matters: For the first time, a nonprofit organizer with a clear problem and two weeks of focus can ship a real tool. A volunteer dashboard. A case intake form. A phone-bank that routes calls based on your state's voter file. Custom software used to require a developer hire or a five-figure vendor contract. It doesn't anymore.

But the opportunity lives or dies on institutional permission. Staff can't deploy a tool their org hasn't sanctioned. Data policies written for a 2015 tech stack quietly block what 2026 tools make possible. The orgs that adapt will ship tools that fit their actual work. The ones that don't will keep paying enterprise vendors for software designed for someone else's problem.

Source: TechCrunch, April 18

---

Dairy Queen is rolling out AI drive-thru workers across the US and Canada

What happened: Dairy Queen announced it's deploying AI voice agents from a company called Presto to handle drive-thru ordering at locations across North America. The AI is designed to speed up service and, per Presto's own marketing, increase "upselling conversion." DQ joins McDonald's, Wendy's, Taco Bell, White Castle, and a growing list of chains replacing front-line workers with conversational AI.

Why this matters: Food service employs millions of workers in the United States, disproportionately workers of color, immigrants, and young people. The "efficiency" frame is cover for wage suppression. Why raise drive-thru pay when you can swap the worker for a voice model that never asks for a raise? And this is happening with almost no public input. Municipalities aren't running AI impact assessments before these systems go live. State legislatures aren't tracking it. Most of us only find out it happened when we pull up to the drive-thru and a different voice takes our order.

There's a progressive labor lane here that almost nobody is running in. Worker voice in AI deployment shouldn't be a luxury reserved for knowledge-worker unions. It should be a baseline demand for any sector getting restructured by this technology.

Source: The Verge, April 18

---

Cerebras filed for IPO, and AI infrastructure is becoming a consolidation story

What happened: Cerebras, an AI chip startup built around massive wafer-scale processors, filed for an IPO on the strength of a run of high-profile deals, reportedly including a major OpenAI contract and a new AWS distribution arrangement. The filing positions Cerebras as one of a small group of companies with the compute capacity to serve frontier AI training workloads.

Why this matters: The companies that own the compute own the roadmap. AI infrastructure is consolidating around a small group of players with deep hyperscaler relationships: Nvidia, Cerebras, Broadcom, AMD, and the custom silicon programs inside Google, Amazon, and Microsoft. No public option. No municipal compute cooperative. No antitrust pressure on the vertical integration between model companies and their chip suppliers. This is how every other technology consolidation has gone, and progressives have mostly watched it happen quietly because we haven't treated AI as infrastructure worth fighting over.

It is. If AI becomes as important to our clients and communities as we think it will, the question of who owns the rails matters as much as the question of which tools we use.

Source: TechCrunch, April 18

---

---

Put AI to work

Practical ways progressives can use AI this week

Upgrade your org's cybersecurity with AI tools

Progressive nonprofits are high-value targets. Voting rights groups. Reproductive health orgs. Immigrant rights legal services. Tenant unions. Labor organizations. The data you hold is exactly what hostile state AGs, litigants, and actual attackers want. Enterprise security tooling used to require enterprise budgets. That's changing fast, and AI is the reason.

Last week Anthropic unveiled Claude Mythos Preview, a new cybersecurity-focused model, and early reports suggest it's thawing the company's relationship with the Trump administration. The deeper story is that every major AI player is shipping security tooling right now, and the nonprofit pricing tiers are finally catching up. If your security posture was built in 2019, this is a real opening.

Tools worth evaluating this week (all have real nonprofit pricing or free tiers):

• Microsoft Defender for Business. AI-driven endpoint protection, included in Microsoft 365 Business Premium for Nonprofits.
• Google Chronicle Security Operations. AI-triaged SIEM. Available through Google for Nonprofits pathways.
• Anthropic Claude Mythos Preview. New cybersecurity-specialized model; request access and evaluate for your threat profile.
• Huntress. Managed detection and response built for teams under 500 staff.
• 1Password + Kolide. Device trust and secret management with nonprofit pricing.

Steps to take this week:

1. List the sensitive data your org actually holds. Donor records, client case files, legal matters, voter contacts, internal board documents. Be specific. Generalities hide the real risks.
2. Map what protects each category. Endpoint antivirus, multi-factor auth, email filtering, backups, disk encryption, access logs.
3. Identify the weakest link. There's always one.
4. Get one demo from a nonprofit-friendly AI security vendor by Friday.
5. Bring the findings to your next ops meeting with a concrete recommendation.

Your real AI-era risk has nothing to do with the tools taking your job. It's that your 2019-era security posture will leak donor records or client data while the org is still debating whether to pay for ChatGPT. Fix the thing that will actually hurt people first.

---

---

Looking ahead

The question inside progressive institutions isn't "should we use AI." That one's settled, whether leadership has caught up or not. Your staff decided. The live question is whether your org catches up on purpose: policies that protect sensitive data, approval paths that let staff actually ship tools, bargaining rights that put workers at the deployment table, security budgets that belong in 2026 instead of 2019.

The people we're fighting for aren't going to thank us in 20 years for holding moral high ground while the best tools available sat on the shelf. What they'll remember is whether we picked the tools up and used them well.

That's the permission gap closed. That's the work.

Until next time,
Jordan

---

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

Spotlight · News · Put AI to Work · Looking Ahead

Creator spotlight: Sky Speirs

@skyspeirs on TikTok, "Solving the AI Power Problem"

Sky is an energy and climate educator who posts short videos about how AI actually gets powered, cooled, and built. Who pays for it. What the alternatives look like. Why "AI is bad for the planet" keeps missing the real fight. It's the kind of translation work the public conversation has been starved of.

A few of her hooks from the current playlist:

• "Who should pay for AI energy costs: you or big tech?" Texas just passed a law requiring large energy users to pay minimum transmission charges. If the largest users covered the state's $33 billion grid investment, the average household would save $200 a year.
• "Will AI break the grid? Short answer: No." The argument: AI doesn't have to be an energy hog. With flexibility, data centers can consume power when clean energy is abundant and ease off when it's not. "We have the headroom, we have the tech, we have the code, and we just need archaic rules to change."
• "Data centers don't need to use a lot of water." Most facilities use evaporative cooling because it's cheap, not because it's good. Liquid cooling cuts water use by up to 90%. "Water-intensive data centers aren't a given. They're a design problem."

Sky's refrain across the series: "We created the problem so we can fix it." That's about as PFAI a sentence as exists on the internet.

Watch the playlist. If you do community education on AI, climate, or energy policy, her videos are a ready-made resource. Send them to your volunteers. Share them in your coalition Slack. Quote her at your next planning meeting.

---

AI News Roundup

A Vermont newsroom just wrote the AI union contract other locals should copy

What happened: On April 1, the VTDigger Guild ratified a four-year contract with the Vermont Journalism Trust that includes some of the most concrete AI protections yet negotiated in a small-newsroom unit. Among the provisions: management must give 60 days' notice before deploying any new AI system, the union has the right to negotiate over effects, laid-off workers affected by AI-driven restructuring get a minimum of 12 weeks of severance, and reporters can withhold their bylines from any story where AI tools played a substantive role. The contract also includes explicit language that "generative AI tools do not adequately substitute for human judgment" in journalism.

Why this matters: VTDigger is not a giant bargaining unit, and that's exactly why the contract matters. Small units often get held up as places where AI protections are impossible to negotiate. Here's a template that proves otherwise. Every clause is specific, enforceable, and portable. Notice periods, effects bargaining, severance floors, byline rights. Pick the ones that fit your shop and bring them to your next negotiation.

---

Lawmakers are quietly using AI to draft bills, without telling anyone

What happened: Transformer News reported this week that state and federal legislators are increasingly using AI-powered drafting tools from companies like FiscalNote and Vulcan Technologies to write and analyze legislation. Virginia and Vermont have both started piloting these tools. The catch: there are virtually no disclosure requirements. Lawmakers can use AI to draft bills, summarize hearings, and analyze amendments without ever telling constituents, colleagues, or opponents.

Why this matters: This is a transparency fight with a clean progressive frame. The tools themselves aren't the problem. AI can make understaffed legislative offices faster and more effective, which is good for democracy. The problem is equity. If premium AI drafting tools cost $50,000+ per seat and corporate lobbyists use them to shape legislation while under-resourced advocacy groups can't, the playing field tilts further toward the people who already write the rules.

---

In brief

OpenAI is buying a personal finance startup. OpenAI announced this week it's acquiring Hiro, a personal-finance AI startup, signaling that financial-planning features are headed to ChatGPT. Before that rolls out, someone needs to answer a basic question: what standard of care does an AI chatbot owe to a user asking where to put their retirement savings? Human financial advisors operate under fiduciary rules for a reason. If an AI tool is going to give financial advice to hundreds of millions of people, the advice should meet the same standard. Push your state financial regulator to start the conversation.

---

---

Put AI to Work

Practical ways progressives can use AI this week

Ask AI what your website is actually doing

Most progressive organizations have Google Analytics installed and almost never look at it. The reports feel intimidating, the interface is dense, and nobody has an afternoon to spare. AI can fix that in about 20 minutes.

Step 1: Pull three reports from GA4. Log in to Google Analytics, go to Reports in the left sidebar. Export these three as CSV (the share icon in the top right of each report, then Download File → CSV):

• Reports → Life cycle → Engagement → Pages and screens
• Reports → Life cycle → Acquisition → Traffic acquisition
• Reports → Life cycle → Engagement → Events

Set the date range to the last 30 days before you export.

Step 2: Upload and ask the big question. Start a new conversation with Claude or ChatGPT. Upload all three CSVs. Then paste something like this:

I run [a progressive nonprofit / a campaign / a small advocacy group] at [URL]. My main goals are [e.g., newsletter signups, event RSVPs, donations, calls to action]. I've attached the last 30 days of Google Analytics data. Tell me: (1) What's working that I should do more of? (2) What's underperforming and why? (3) What are three specific changes I should test next month?

Step 3: Follow up. Once you have the answer, keep going. Good follow-ups:

• "Which pages convert visitors into [your goal]? Which don't?"
• "Am I losing traffic on mobile vs. desktop? What might be causing it?"
• "What content themes should I make more of based on what people are actually reading?"
• "If I only had time to fix one thing on this site, what would have the biggest impact?"

Bonus: use AI to set GA4 up properly. If your GA4 isn't actually tracking the things you care about (signups, RSVPs, donations), take a screenshot of your GA4 admin panel and ask Claude to walk you through configuring events and conversions for your specific goals. It'll give you step-by-step instructions with the exact buttons to click.

AI isn't replacing your data analyst here. It's giving a busy organizer enough of an analyst to make better decisions this month. Progressive orgs don't need more tools. They need more capacity, and this is a free way to get some.

---

---

Looking Ahead

Tuesday, April 20: California Senate floor vote on SB 947, the No Robo Bosses Act. If you live in California, this is the week to make noise. If you don't, this is the week to start pitching your own state legislators on a companion bill. Progressive policy wins are built on templates. Someone has to go first. It might as well be California.

And if you're looking for a way to engage that doesn't involve another news cycle: go watch Sky Speirs. Then go send her playlist to someone who thinks AI is either going to save the world or end it. The real answer is harder and more interesting, and our side has the better version of that answer.

Until next time,
Jordan

---

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

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

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

Quick Take

We've been saying since issue one: the answer to bad AI isn't less AI. It's enforceable rules, human oversight, and showing up with real ideas.

This week, that argument landed in a courtroom. A federal judge blocked the Pentagon from blacklisting Anthropic over its refusal to drop ethics guardrails on military AI. She called it "classic illegal First Amendment retaliation." Meanwhile, the White House AI czar position is vacant with no replacement planned, which means the real AI regulation fight is at the state level, where progressives have the most leverage right now.

Let's get into it.

---

AI News Roundup

There's no one steering federal AI policy. State legislatures are where you should be showing up.

Photo by Brett Sayles / Pexels

What happened: David Sacks, the venture capitalist Trump appointed as Special Advisor on AI and Crypto, hit his legal 130-day term limit and moved to an advisory role. The administration says it won't appoint a replacement. There is no longer a White House AI czar.

Sacks was behind the push to preempt state-level AI protections that we covered in issues four and seven. With his departure, the administration's AI agenda loses its public champion. But the real action was already moving to state legislatures, and the pace is accelerating.

Since we started this newsletter in January, Washington passed five AI bills in a single session, including transparency requirements for AI in health insurance decisions. Oregon's chatbot safety bill passed 52-0. California has two workplace AI bills moving with bipartisan support. Virginia advanced an AI regulation bill 39-1. Colorado's AI Act is in effect. Montana wrote the right to use AI into law. And we tracked in issue six how New York is debating whether AI should be allowed to answer questions in licensed professions, with progressives on both sides pushing for thoughtful access, not gatekeeping.

Federal AI policy is stalled or hostile. State policy is where the real work is happening, and it's where progressive voices can have the most impact right now.

---

TikTok and Samsung showed, again, why voluntary AI commitments are worthless

Photo by Bastian Riccardi / Pexels

What happened: Samsung ran AI-generated video ads on TikTok without the disclosures required by TikTok's own policies. The exact same videos were labeled as AI-made on YouTube, where Google actually enforces its rules. The kicker: both TikTok and Samsung are members of the Content Authenticity Initiative, an industry group that exists specifically to promote AI transparency standards.

This isn't a one-off. Companies join voluntary AI transparency initiatives, put out press releases about responsible AI, and then ignore their own rules when nobody's watching. YouTube labels the same ads because it has enforcement mechanisms. TikTok doesn't because it doesn't.

Why this matters: We've been beating this drum since issue three: voluntary AI commitments evaporate without enforcement. Anthropic's ethical position held this week because binding legal standards backed it up. TikTok's transparency policy collapsed because nothing did. Every story we've covered points the same direction: companies do the right thing when the rules require it. Not before.

---

In brief

A federal court backed enforceable AI standards, and it matters. Judge Rita Lin granted Anthropic a preliminary injunction against the Pentagon's attempt to blacklist it as a "supply chain risk" for maintaining ethics guardrails on military AI. Her 43-page ruling called the designation "Orwellian" and the blacklisting "classic illegal First Amendment retaliation." The Pentagon CTO's office says the ban "still stands" despite the ruling, setting up a potential contempt fight. The takeaway: voluntary promises are rewritable. (Anthropic itself rewrote its safety pledge weeks earlier.) Court orders aren't. The progressive case for enforceable rules just got 43 pages of federal case law backing it up.

---

---

Put AI to Work

Practical ways progressives can use AI this week

Build a custom issue monitoring feed

If your organization tracks policy issues, you know how this goes. Searches pile up. Google Alerts send you junk. Important stories slip through because they didn't use the right keywords.

Here's a practical setup that takes about 30 minutes and uses free tools:

Use Bluesky's custom feeds. Bluesky's open protocol means anyone can build algorithmic feeds, no corporate gatekeeper. The Bluesky Feed Creator lets you create keyword-based feeds without writing code. Build one for your issue area: "housing justice" + "AI" + "automation," or "reproductive rights" + "legislation" + "2026." Pin it and check it daily alongside your main timeline.

Set up RSS monitoring with AI filtering. Most news sites still publish RSS feeds. Use a free reader like NetNewsWire (Mac/iOS) or Feeder (web/Android) to subscribe to outlets that cover your issues. For higher-volume feeds, paste your latest batch of headlines into Claude or ChatGPT and ask it to "flag the 3 most relevant stories for an organization focused on [your issue], and explain why each one matters for our work."

Create a weekly digest prompt. At the end of each week, paste your collected bookmarks and feed items into an AI with this prompt: "You're a policy analyst at a progressive advocacy org focused on [issue area]. Review these articles and produce a one-page briefing: what happened this week, what our supporters need to know, and one action item we could promote." That output can go straight into a staff Slack channel, board update, or supporter email.

The whole system is free and replaces hours of manual scanning each week. That's time back for actual organizing.

---

---

Looking Ahead

I keep coming back to a line from issue two: "What if regulating AI actually makes AI better?"

This week gave us a real answer. A federal court said the Pentagon can't punish a company for maintaining ethical red lines. States are passing AI bills with bipartisan support while the federal government's AI chair sits empty. And the SPLC's own workers bargained for AI protections and won. Enforceable standards create the trust that makes adoption possible. Without them, you get a race to the bottom where the company willing to do the most reckless thing wins.

Progressives have been making this argument about every other industry for a century. Clean air regulations didn't kill manufacturing. Seatbelt mandates didn't stop people from driving. Rules make systems work better for more people.

AI is no different. And this week, a federal judge agreed.

Until next time,
Jordan

---

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

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

Quick Take

Here's a thing that's actually happening right now: doctors at Kaiser Permanente are looking their patients in the eye again. AI scribes (tools that listen to the conversation and write the clinical notes afterward) saved Kaiser's physicians the equivalent of 1,794 full workdays of documentation time last year. Doctors report getting back about an hour a day they used to spend typing into a screen. Patients say the visits feel more human.

That's a genuine, measurable good.

Here's the other thing that's happening right now: health insurance companies are deploying AI to process prior authorization requests — the step between your doctor ordering something and your insurer agreeing to pay for it. Less human review. Faster denials. Same underlying technology.

AI helping your doctor listen to you. AI helping your insurer ignore you. Same technology, opposite purposes.

That's the whole AI fight in one frame. The question was never "is AI good or bad?" It was always "who's pointing it at whom, and who gets to decide?" States are starting to answer that question. Washington just passed five AI bills in a single legislative session, including one that forces insurers to report how they're using AI to deny claims. But eight days later, the White House released a framework designed to wipe out every state-level AI protection in the country.

Let's get into it.

---

AI News Roundup

AI is giving doctors back their evenings. It's also helping insurers say no.

Photo by Vitaly Gariev / Unsplash

What happened: A wave of peer-reviewed studies published in the past year shows AI clinical scribes are delivering real results. A University of Wisconsin randomized trial found providers saved 30 minutes a day and saw measurable drops in burnout. A UCLA trial across 14 specialties found a 7% improvement in burnout scores, with fewer than 10% of patients declining to use the technology. At UI Health in Chicago, patient satisfaction scores for "my provider explained things clearly" jumped from 91% to 97% after AI scribes were deployed.

The context matters: doctors currently spend nearly half their workday on electronic health records and desk work, and almost two hours every evening catching up on documentation after their kids are in bed. Nearly half of primary care physicians are burned out. AI scribes are making a specific group of people's lives measurably better.

But the same AI capabilities that help doctors are being deployed very differently by health insurers. AI-powered prior authorization systems can process claims faster, which in practice often means denying them faster, with less human review. That's why Washington passed SB 5395, requiring health insurers to report quarterly on how AI is involved in claim decisions.

Why this matters: This is the clearest illustration of why blanket AI regulation — whether "ban it all" or "deregulate everything" — misses the point. AI helping a burned-out doctor spend more time listening to you is good. AI helping an insurer process your denial without a human ever looking at it is bad. The technology is the same. The difference is who it serves and whether anyone's watching.

Washington's approach (don't ban the technology, require transparency about how it's being used) is the kind of regulation progressives should be championing everywhere. And it's exactly the kind of state-level protection that the federal preemption framework (see below) would override.

---

States are passing AI protections. The White House wants to wipe them out.

What happened: Before adjourning on March 12, Washington's legislature passed five AI bills in one session: content disclosure, chatbot safety for kids, restrictions on AI in health insurance prior authorizations (SB 5395), bans on AI-generated child sexual abuse material, and property rights in digital likenesses. Oregon's chatbot safety bill passed 52-0 in the House. California has two workplace AI bills moving through committee with bipartisan support.

Then on March 20, the White House released a national AI framework directing Congress to preempt state AI laws. The administration calls it preventing "a patchwork of fifty discordant rules." The framework calls for no new oversight body, reduced liability for AI companies, and shifts child safety responsibility from platforms to parents. House Republican leadership immediately backed it.

Why this matters: Last issue, we noted that legal analysts think state AI protections are more durable than anyone expected — executive orders alone probably can't preempt state law. That's still true. But this framework isn't an executive order. It's a roadmap for Congress to do the preempting instead. If it works, every state-level protection that progressives spent years building — Colorado's AI Act, Illinois' disclosure requirements, California's worker protections, and now Washington's health insurer transparency rules — gets overridden by a single federal standard written with industry input.

The states that moved fastest are proving that AI regulation works and has bipartisan support. Oregon's chatbot bill didn't pass 52-0 because it was controversial. Washington's health insurer transparency bill passed because people understand what it means when a computer denies your insurance claim. That real-world track record is exactly what the preemption push is trying to short-circuit.

---

The union had negotiated AI-specific contract language before any of this happened. When POLITICO deployed the tools anyway, the union filed a grievance. The arbitrator ruled that POLITICO violated the collective bargaining agreement and ordered a 60-day bargaining period with a negotiated remedy.

Here's why this story matters: corporate AI ethics policies are voluntary — companies can change them whenever they want (see: every AI safety pledge that's been quietly dropped in the past year). Contract language is enforceable. The CWA now has 58 newsroom contracts with AI provisions. ZeniMax workers at Microsoft negotiated the first video game industry AI contract. Frontier Communications workers in California won a requirement for joint deliberation before any AI tool gets deployed.

The pattern: workers who wrote the rules before AI arrived had leverage. Workers who didn't are catching up.

---

Put AI to Work

Practical ways progressives can use AI this week

Field organizing tools built in a weekend

The Cooperative Impact Lab ran a generative AI hackathon with progressive organizations last year. Two of the tools that came out of it are worth knowing about, not because they're polished products, but because they show what a small team can build in a couple of days when they understand their community's actual needs.

Fair Count's "Community Voice" tool started with a simple insight: canvassers know things that databases don't capture. Fair Count, a Georgia-based census and voting rights organization, had canvassers in Mississippi record 30-to-90-second voice memos right at the doorstep after each conversation. They collected 120 of these and ran them through AI transcription and sentiment analysis.

What came out: county-level breakdowns of what people actually care about. Voting intention mapped by neighborhood. Pre-canvassing "topic primers" so the next organizer who knocks on that door walks in prepared. The kind of strategic intelligence that well-funded campaigns pay consultants six figures for — built by hackathon participants in a weekend.

Dr. Jeanine Abrams McLean, Fair Count's president: "The fact that hackathon participants were able to create a functioning tool in that amount of time was really mind-blowing."

AAPI Victory Alliance's "Truth Tea" tackles a different problem: disinformation targeting communities in languages that English-speaking rapid response teams can't monitor. The tool identifies political disinformation in video content, tested first with Hindi-language material, analyzes the narratives, and generates shareable counter-messaging in the target language. Thirty-four percent of AAPI voters have limited English proficiency. A two-person comms team can't monitor Hindi and Tagalog and Mandarin social media feeds manually. AI can.

If you want to explore what's out there: The Higher Ground Labs AI Resource Guide (updated March 15) is a solid collection of vetted AI tools for progressive campaigns and advocacy organizations, organized by use case with case studies and prompt templates. The AI Campaign Stack directory is community-powered and growing.

---

---

Looking Ahead

The AFL-CIO's inaugural Workers First AI Summit is this Thursday, March 26, in Washington. AFL-CIO President Liz Shuler is delivering the opening keynote, with MIT's Max Tegmark giving the lunch keynote. Rep. Ro Khanna and Randi Weingarten from the American Federation of Teachers are on panels. The summit centers the AFL-CIO's demand for enforceable AI guardrails, worker inclusion in how AI gets deployed, and protection from algorithmic surveillance and displacement.

This is the biggest organized labor response to AI to date. And it's happening the same week the White House is trying to strip away state-level protections.

The through-line of everything in this issue: the people who showed up prepared got better outcomes. Washington legislators who passed AI bills before the preemption push have a track record to defend. POLITICO journalists who negotiated AI contract language before management deployed AI tools had legal standing to fight. Kaiser doctors who adopted AI scribes are spending more time with patients. Fair Count organizers who experimented with voice memos during a hackathon have field intelligence their competitors don't.

The people who waited got AI pointed at them by someone else.

This is still early. The tools are still being built, the laws are still being written, and the people who show up now get to shape what comes next. If you're reading this newsletter, you're already closer to the front of this than most people in our movement. Use that.

Until next time,
Jordan

---

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

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

Quick Take

Harper Carroll, a Stanford CS grad, former Meta AI engineer, and AI educator, posted a video last week about a New York State Senate bill that would ban AI from giving "substantive responses" in medicine, law, engineering, and 11 other licensed professions.

Her argument: the bill doesn't define "substantive." That's not an oversight. It's a feature. Vague language means anyone can sue, which means AI companies will overcorrect and restrict access to be safe. The people who lose first? The single mom checking her kid's symptoms at 11pm. The renter trying to understand a predatory lease. The worker who can't afford $500 an hour for a lawyer.

As Carroll put it: "This knowledge has always been out there, but now it's finally democratized. And New York wants to take that away."

This is the kind of fight progressives should be leading. We don't love AI uncritically. But gatekeeping shouldn't be the answer. The response to AI's real problems (hallucinations, bias, corporate consolidation) is not to hand the keys back to the professions that locked people out in the first place. It's to build guardrails that protect people without recreating the access barriers that harmed them.

I wrote about a version of this on my blog this week: why progressives can't afford to sit out the AI revolution. The short version: 39% of campaign professionals haven't used AI for content creation at all. We can't afford to be in that 39%. We don't get to unilaterally disarm.

Let's get into it.

AI News Roundup

New York wants to ban AI from answering your questions

Photo by National Cancer Institute / Unsplash

What happened: New York Senate bill S7263, sponsored by Senator Kristen Gonzalez, would prohibit AI systems from providing "substantive responses" in medicine, law, engineering, and 11 other licensed professions. The bill is on the Senate floor calendar as of late February. Liability falls on deployers, not model makers, and it creates a private right of action, meaning anyone can sue.

The central problem: "substantive" is never defined. Bloomberg Law flagged this as the bill's fatal flaw. Operators will over-restrict AI outputs far beyond what the statute requires just to avoid lawsuits. There's no clear line between "What is ibuprofen?" and "What dosage should I take?"

Why this matters: The bill is framed as consumer protection, but ask who it actually protects. The people with the most to lose from restricted AI access are the people who can't afford professional gatekeepers. AI is not replacing doctors or lawyers. It's giving people information that was always available but locked behind $500-an-hour consultations. A renter facing a predatory landlord who needs to understand their rights. A worker trying to figure out if their severance offer is fair.

If AI gives bad medical information, the fix is accuracy standards and disclosure requirements. Not banning it from answering the question.

Montana just made AI access a right

What happened: Montana became the first state to write the right to use AI and computational resources into law. Governor Gianforte signed SB 212, the "Right to Compute Act," which says that any state restriction on AI use must be "demonstrably necessary and narrowly tailored to fulfill a compelling government interest." The law also requires mandatory shutdown mechanisms and annual risk assessments for AI systems that control critical infrastructure.

Why this matters: This came from a Republican governor in a deep red state. The bill was championed by libertarian-leaning tech advocates. Progressives should be paying attention, not because the politics align, but because the framework does.

Montana's approach: default to access, require narrow justification for any restrictions, build in safety for critical systems. That's closer to what thoughtful progressive AI policy should look like than most of what our side has proposed. The contrast with New York couldn't be sharper. One state defaults to restriction, the other to access.

The lesson is not that Montana got everything right. It's that progressives are at risk of ceding the "AI access" frame to the right. Access to knowledge and tools should be a progressive value.

Put AI to Work

Practical ways progressives can use AI this week

Use AI to research who's funding your opponents

Photo by Green Chameleon / Unsplash

Opposition research used to require a dedicated researcher and weeks of digging. AI changed that. A two-person advocacy shop can now do solid oppo research in an afternoon.

Map funding networks with 990 filings

Every nonprofit and foundation files a 990 tax return, and they're all public. ProPublica's Nonprofit Explorer has millions of them searchable for free.

1. Look up the think tank or advocacy group opposing your issue
2. Download their 990s (ProPublica has them as PDFs)
3. Upload the PDFs to Claude or ChatGPT and ask: "List every grant this organization received over $50,000, who it came from, and the stated purpose"
4. Then ask: "Which of these funders also fund organizations working on [your issue area]?"

What used to take days of reading tax filings takes 20 minutes. You'll see the funding web behind the opposition before your next coalition call.

Research lobbying connections before hearings

Before a legislative hearing, you want to know who's lobbying on the other side. OpenSecrets tracks lobbying disclosures and campaign contributions, and your state may have its own disclosure database as well.

1. Search for the company or industry group testifying against your bill
2. Copy their lobbying disclosure data into an AI
3. Ask: "Summarize this organization's lobbying spending, which legislators they've contributed to, and what issues they've lobbied on in the past 3 years"
4. Use the results to prep counter-arguments and questions for friendly legislators to ask

Build a "who funds this?" one-pager

For any upcoming meeting, hearing, or media hit:

1. Gather everything you can find about the opposition group (website, leadership, 990s, lobbying records, news coverage)
2. Feed it all into an AI and ask: "Create a one-page brief on this organization: who funds them, who leads them, what positions they've taken, and what conflicts of interest exist"
3. Print it out for your team

This is exactly the kind of work that well-funded organizations have always done and grassroots groups couldn't afford. AI makes it possible for a small shop to walk into a hearing as prepared as anyone in the room.

Looking Ahead

Craig Mod, a writer and photographer based in Japan, published an essay this week called "Software Bonkers." He built custom accounting software in five days using AI coding tools. He's not really a programmer. He just needed something that fit his multi-country, multi-currency freelance life, and nothing off the shelf did.

His line that stuck with me: "The software feels organic and pliable in a form perfectly shaped to my hand, able to conform to any hunk of data I throw at it. It feels like bushwhacking with a lightsaber."

That's the access story applied to software itself. Custom tools used to require hiring developers or buying enterprise licenses. Now a nonprofit organizer or a freelance consultant can have software shaped to their exact workflow. That's happening right now, to regular people.

New York says only licensed professionals should answer your questions. Montana says everyone has the right to compute. The reality we should be fighting for is somewhere specific: access with accountability. Guardrails that protect people without gatekeeping them out.

That's the progressive position on every other technology. It should be ours on AI too.

Until next time,
Jordan

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

From our friends

Your org deserves its own AI. Not Big Tech's.

Change Agent is a private AI platform built for nonprofits, unions, and advocacy orgs. Your data stays yours, it plugs into tools you already use (Google Drive, Slack, ActBlue), and it handles the tedious stuff so your team can focus on the mission. Starts at $35/month. Small nonprofits under $1M can apply for discounted pricing.

Learn more