Bernie wants you to own a piece of AI

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

In this issue

Bernie Sanders wants the public to own a piece of the biggest AI companies, and put the idea in the New York Times.
Congress still can't read its own AI homework, so the companies are writing it. There's a fix, and it's a door progressive orgs can walk through right now.
A clear, copyable template for how states should handle AI in K-12 schools, built around adoption with guardrails instead of bans.
GLAAD's new report makes the case that building AI to include everyone makes it work better for everyone.
Progressive AI Win: AI helped doctors crack 18 rare disease cases that had gone unsolved for kids the system had given up on.
Put AI to Work: build a credible two-page AI briefing for your member of Congress in an afternoon.

Quick Take

Most of the AI conversation gets stuck on one verb: regulate. Set the rules, slow it down. All of that matters. But this week a different verb showed up, in of all places the New York Times opinion page: own.

Bernie Sanders put public ownership of the biggest AI companies on the national table, and the argument underneath it is one we have been making here since the start. AI was built on all of our work, so the gains from it should come back to all of us. That is a more hopeful frame than "make it stop," and it changes what you ask for. You stop asking only for guardrails and start asking for a stake.

The rest of the issue follows that thread: who owns AI, who gets briefed about it, and who gets built into it. Let's get into it.

$48.5 million

That's what just four companies (Anthropic, OpenAI, Meta, and Google) spent lobbying the federal government in 2025. Meanwhile Congress has almost no AI expertise of its own, so when a staffer needs to understand a bill, the most fluent voice in the room usually works for one of those four. You can't have public power over AI when only one side of the table can afford to show up. Transformer News has the breakdown.

AI News Roundup

Bernie Sanders wants you to own a piece of AI

What happened: In a New York Times op-ed, Senator Bernie Sanders said he will introduce the American AI Sovereign Wealth Fund Act. The mechanism is blunt: a one-time 50 percent tax on the largest AI companies (OpenAI, Anthropic, and xAI), paid not in cash but in stock, with that equity deposited into a public fund. In return the public would get voting rights, seats on the boards, and eventually direct payments to ordinary Americans. His case for it is the same one this newsletter keeps coming back to. As Sanders put it, "AI was not created out of thin air. The foundation of AI is built on our collective human intelligence: our books, songs, artwork, journalism, computer code, scientific research, videos, conversations, images and ideas spanning generations." So, he argues, "the wealth it generates must benefit humanity."

Why this matters: For two years the progressive AI conversation has mostly been defensive: stop the harms, slow the rollout. Those fights are real and worth having. But ownership is a different posture, and a stronger one. It says the public already paid for this, in the form of everything we ever wrote and made and shared, and we are entitled to a return. Here's the part worth noticing: the Trump administration has already taken equity stakes in around twenty private companies, in minerals, semiconductors, and quantum computing. So the government taking a piece of strategic industries is no longer a fringe idea. What's missing from that version is the half that matters most to us, which is the public actually getting the upside. A fund that the same handful of people quietly control is not the goal. Ownership only means something if the people who own it can steer it, and if the dividend actually lands in regular households.

What you can do

The next time AI policy comes up with your representatives or allied policy shops, add the ownership question to the regulation question. Ask who owns the upside, and how it reaches the people whose work built this, not only how to keep it safe. Sanders just made that a serious thing to ask for, and asking for it by name is how it stays on the table. Sources: Fortune, Washington Times.

Congress can't read its own AI homework, so the companies are writing it

What happened: A piece in Transformer News lays out a problem that quietly shapes every AI bill: Congress has very little in-house technical expertise on AI, so staffers lean on briefings from the companies themselves. The fixes on the table are practical, not exotic. Rebuild the Office of Technology Assessment, the in-house science shop Congress shut down in the 1990s. Expand the Government Accountability Office's tech team, which has already grown from 49 staff to more than 100. Fund more committee staff. And back fellowship programs like TechCongress that drop technical experts straight into congressional offices.

Why this matters: This is the same thread as the lead story, one rung down. You can't have public power over AI if the public's own representatives only hear about it from the people selling it. One reformer in the piece put it plainly: "Why are we leaning on the tech private sector to supply brain power to the legislative branch when the legislative branch can just hire it itself?" The good news is that this is a fixable, fundable gap, and it's a wide-open door for progressive policy organizations. The fellowships already exist. The offices are short on help. A progressive technologist placed in the right office shapes how a staffer understands AI for years.

What you can do

If your organization has anyone with policy or technical chops, look up the AI and tech fellowship programs (TechCongress is the best-known) and share them with your team. Placing one of your people inside a congressional office does more to shape AI policy than a dozen sign-on letters. Source: Transformer News.

A clean template for AI in schools

What happened: The Center for Democracy and Technology published three policy priorities for states writing laws about AI in K-12 schools. The first is risk management: name the high-stakes uses of AI in a school, and require real protections around student data, including human review before and after a system goes live. The second is AI literacy: teach students and staff how to use these tools responsibly, including how to spot a biased or flat-out wrong answer. The third is community engagement: build model policies that cover privacy, security, deepfakes, and discrimination, and give parents and community members a real way to weigh in as the policies develop.

Why this matters: Schools are where a lot of families meet AI policy for the first time, usually through a panicked headline about cheating. CDT's framing is the opposite of panic. It assumes the tools are coming and asks how to bring them in well, which is exactly the shape of argument this newsletter favors. And because it's written for state legislation, it's portable. A school board member, a state legislator, or an education advocate can pick it up and have a starting structure without building one from scratch.

What you can do

If you work with schools, parent groups, or education advocates, send them CDT's three priorities as a checklist. It turns "we should do something about AI" into three concrete questions a school board can actually answer. Source: CDT.

GLAAD: building AI for everyone makes it better for everyone

What happened: GLAAD released its first AI report, "Build for Everyone," a framework for LGBTQ representation and safety across the whole life of an AI system. It walks the full pipeline, from how models are trained to how they get deployed, moderated, and eventually handed agent-like autonomy. The recommendations are build directions, not bans: fix biased training data at the foundation, don't let automated systems quietly discriminate, keep humans in the loop, collect less personal data and protect what you do collect, and bring civil society in from the start. The report grounds it in real harm, including a case study on AI systems generating misinformation about conversion "therapy," a discredited and dangerous practice.

Why this matters: The usual story about inclusion in tech treats it as a cost, a nice-to-have you bolt on if there's budget. GLAAD flips that. Their argument is that a model trained on thin or skewed data about a group of people produces worse, less reliable answers for everyone who asks about them, and fixing that makes the product better across the board. That reframing matters well beyond LGBTQ users. It's the same logic that should drive how AI handles disability, language, race, and faith. Accuracy and inclusion are the same project.

What you can do

If your organization is evaluating or buying an AI tool, use GLAAD's framework as a vendor checklist. Ask how the model was trained, what human oversight exists, and what data it keeps. Those questions protect the communities you serve, and they happen to surface the better products too. Source: GLAAD.

Progressive AI Win

AI helped crack 18 cases the system had given up on

OpenAI, Boston Children's Hospital, and researchers at Harvard pointed an AI research model at 376 rare pediatric disease cases that had gone unsolved, some for years. The model read through de-identified clinical notes, genetic data, and symptom records, and surfaced possible diagnoses for human specialists to check. The result: 18 new diagnoses and 7 rediscoveries, confirmed by doctors using the standard clinical process.

Sit with who those patients are. Children with conditions so rare that no single doctor could hold the full pattern in their head, in a corner of medicine where there was never enough profit to justify the search. The AI didn't replace the specialists; it gave them a lead they didn't have, and they did the rest. That's the version of this technology worth fighting for, aimed squarely at people the system had already written off. Source: The Neuron.

Put AI to Work

Practical ways progressives can use AI this week

Brief your member of Congress in an afternoon

The Congress story above isn't just news, it's an opening. Offices are hungry for plain, credible information on AI, and right now they mostly get it from the companies. You can change that for one office this week, even without a policy analyst on staff. Here's a workflow to produce a sharp two-page briefing memo on one AI issue your organization cares about.

1. Scope it in Perplexity (about 20 minutes). Ask it to find recent policy developments, agency guidance, and credible advocacy positions on your specific issue (say, AI in benefits eligibility, or AI in tenant screening). Prioritize 2025 and 2026 sources. Collect five to seven links.

2. Draft it in Claude or ChatGPT (about 30 minutes). Tell it: you're writing a two-page briefing memo for a congressional staffer with no technical background; here's the issue, here's the community it affects, here are the sources. Ask for a plain-English explainer, three things happening in this district or state, two concrete asks, and one note on what other offices are doing. No jargon.

3. Plain-language pass (about 20 minutes). Run the draft through the Hemingway App and aim for an eighth-grade reading level. Cut anything it flags as hard to read. Staffers read fast.

4. Add one local number by hand (about 15 minutes). Look up a single statistic specific to the district, the number of Medicaid enrollees, local eviction filings, whatever fits. That one detail is what makes a staffer forward it up the chain.

5. Send it. Email the chief of staff or legislative director, attach it as a PDF, and offer a short follow-up call. Subject line: "Brief on AI and [your issue], two pages, local context included."

The whole thing fits in an afternoon, and it puts a credible progressive voice in front of an office that would otherwise hear only from the vendors. That's the knowledge gap from the news section, closed one memo at a time.

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

Looking Ahead

Three things to keep an eye on.

Watch the Sovereign Wealth Fund Act. The real test is whether it picks up co-sponsors and a bill number, or stays an op-ed. Either way, "own it" is now part of the vocabulary, and that's the win to build on.

Watch which states pick up the K-12 framework. CDT wrote it to be copied. The first states to turn those three priorities into law set the template everyone else borrows.

Watch whether any AI company adopts inclusive-design commitments. GLAAD handed the industry a roadmap. The interesting question is who picks it up voluntarily, and who has to be pushed.