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

Quick Take

Three groups of workers got something about AI put in writing over the last few weeks.

SUNY’s faculty union got a clause making a human being answerable for anything AI produces in a course. ProPublica’s newsroom got a transparency mandate, after three years of fighting for a first contract. Songwriters got a licensing framework from Spotify, though not yet the numbers that would tell us whether it’s fair.

Then there’s a fourth group. The people who organize against a data center have no contract, no local, and no institution standing behind them. What they got, according to the American Prospect, was private intelligence firms selling dossiers about critics like them, in one case pitched to a federal regulator.

I don’t think that contrast is an accident, and it’s most of what I want to talk about this week.

Let's get into it.

AI News Roundup

Somebody is selling a file on the people who show up

What happened: The American Prospect reported on August 10 that private intelligence firms are selling dossiers on critics of AI and the data centers powering them, and that at least one of them tried to sell that product to a federal regulator.

The core finding, in the Prospect’s own words: “Private intelligence firms are selling dossiers about critics of artificial intelligence and the data centers powering them, including trying to peddle them to federal regulators, according to documents obtained by the Prospect.”

Three firms are named. According to documents the Prospect obtained, RANE Network offered the Federal Energy Regulatory Commission a report on “anti-technology sector sentiment” in March 2025, and ran a webinar on anti-tech backlash that May. Kroll sells “risk intelligence” covering activism and reputational events to data center operators, which you can read on Kroll’s own site. And Liferaft, a subsidiary of Securitas, published a report on July 31 monitoring online “Threats to AI Infrastructure and Executives.”

Why this matters: I want to be careful here, because there’s an easy version of this story that I don’t think is the right one.

The easy version is that data centers are bad and this proves it. That’s not the argument. We ran an entire issue in May arguing communities should own a piece of these facilities rather than block them, and I still think that’s right. Build them. Regulate them.

The argument is narrower, and it’s mine rather than the Prospect’s. If you are going to build facilities on this scale, you need people willing to show up to the hearings and say what they think. That participation is the thing that makes the permit legitimate. And the surest way to end it is for the people considering whether to speak to learn that somebody sells a product describing them.

Notice who the pitch went to. Not a private operator worried about trespassing, which would be ordinary corporate security. A federal regulator, the agency whose job is to weigh the public interest in energy infrastructure. The sales pitch is that the regulator should understand opposition as a category of risk to be managed rather than as testimony to be weighed. That is a different thing, and it’s the part I’d want a member of Congress asking about.

And look at the matchup. On one side, firms that do this professionally, with money behind them and very little public visibility. On the other, the people who typically turn up to these proceedings, which is to say volunteers and neighbors reading prepared statements into a microphone on a weeknight. Nobody in that second group has a budget line for counter-intelligence.

Which is the thread running through the rest of this issue. Everyone else in here had a bargaining committee behind them. These folks have a sign-up sheet.

42,000 SUNY workers wrote down who’s responsible when AI is in the room

What happened: United University Professions, which represents more than 42,000 members across the State University of New York, ratified a five-year contract that runs through July 1, 2031. Inside Higher Ed covered the ratification on August 4.

The AI language is short and does a lot of work. All SUNY courses stay under the “direction and responsibility” of humans, who hold “ultimate accountability” for work done by AI.

The rest of the contract is what you’d expect from a long fight: annual raises starting at 4.5% this year, expanded due process and job security protections, and less weight on student surveys in evaluating teaching.

Why this matters: Read what that clause does and doesn’t do, because the doesn’t is the interesting half.

It doesn’t ban AI from SUNY classrooms. The reported language sets up no committee, no waiting period, no approved-tools list. What it establishes is that a person’s name stays attached to the outcome, and that person is a human being who can be asked about it.

That’s an unglamorous kind of win and I’d argue it’s the most durable kind. Rules about which tools are permitted go stale in about eighteen months. A rule about who is accountable survives the tool changing, because it was never about the tool.

And it sidesteps the argument higher ed has been stuck in for three years, the one about AI and cheating, which treats all of this as a discipline problem. This contract treats it as a labor question instead. A person’s name goes on the output, and the people whose names those are had a union at the table when the rule got written. Those two facts are related, and I’d argue the second one produced the first.

Briefly — The Spotify deal we covered in May added another publisher

Back in May, Issue 16’s Progressive AI Win was Spotify’s licensing agreement with Universal Music Group for a tool letting fans make AI covers and remixes, built on what the companies called consent, credit, and compensation. I flagged at the time that the revenue split would decide whether it was actually fair.

On August 13, Kobalt signed on. Charlie Hellman of Spotify used the same formulation: “This agreement with Kobalt is grounded in consent, credit, and compensation for the songwriters who take part.” The tool is still coming rather than shipped, and it’ll be a paid add-on for Premium subscribers, with the stated intent of generating additional income for the songwriters involved.

Another publisher in and the terms still aren’t public. That’s the thing to keep asking about. Getting the shape right matters, and I’ll take a framework built to pay creators over one that wasn’t. But at some point somebody has to show us the percentage.

Source: Spotify Newsroom, 13 August 2026

Briefly — No Robo Bosses cleared Assembly Appropriations

Last issue I flagged that California’s SB 947 was heading into Assembly Appropriations, the fiscal gate where bills quietly stop existing. On August 13, it cleared, 10 to 4, and was ordered to third reading.

That means Senator Jerry McNerney’s bill, formally the No Robo Bosses Act of 2026 and sponsored by the California Federation of Labor Unions, AFL-CIO, now goes to a full Assembly floor vote. It would bar employers from relying solely on an automated system to fire or discipline a worker, require human oversight where such a system assists those decisions, and prohibit systems that use a worker’s personal information to predict future behavior.

Sources: California Legislature SB 947 status; Office of Senator Jerry McNerney

Briefly — A new AI advocacy organization is hiring four people

Somebody is standing up a new advocacy organization to build grassroots political power around AI policy, launching publicly in late summer, and they’re hiring now.

Four roles, each with its own application, and all four post a range:

• Organizing Director, $150,000 to $175,000. New York or San Francisco preferred, U.S. remote possible.
• Operations Director/Manager, $130,000 to $170,000. Same location setup. Fully covered health, dental and vision for you and your family, and an automatic 3% 401(k) contribution.
• Senior Digital Manager, $120,000 to $145,000. New York preferred, remote possible for the right person.
• Distributed Organizer, $2,000 a week full-time and prorated for part-time, on a six to eight week contract starting September 1. Remote in the U.S.

Two things to know before you apply. The organization isn’t named on any of the forms, which is normal enough for a pre-launch hire and still means you’re applying somewhat blind. And its stated mission spans economic benefit, personal freedoms, and preventing the development of superintelligence that poses existential risks.

Posting the ranges up front, though, is the right call and worth saying so. Organizing jobs in AI policy are still rare, these pay like real jobs, and a few of you are qualified. The overview page has all four.

Source: hiring page, fetched 17 August 2026

Put AI to Work

Practical ways progressives can use AI this week

Two tools for making things look good with no design budget

Jeremy Caplan wrote up two visual tools on August 15 that are worth a look if your organization makes its own graphics, which is to say most of you.

Whorl is a browser-based design tool for animated graphics built from moving 3D shapes. Posters, social posts, title cards, kinetic text. The thing that makes it different from the prompt-and-pray tools is that you actually manipulate things on a canvas. Caplan’s description: “Rather than just typing in a prompt, you’re actually moving things around on the canvas, making your own digital art.” He uses it for kinetic text graphics in his teaching slides.

It’s free with a watermark. Watermark-free is $5 a month or $80 a year, and there’s a higher-resolution tier at $180 a year. Seven-day trial at design.whorl.app, no card required, and Caplan has a code, WONDER30, for 30% off the first year.

Flux 3 is Black Forest Labs’ video generator: short clips, up to twenty seconds, from a text prompt or an uploaded image. Good for presentations, explainer beats, and story starters.

One correction to anything you may have read last week. It was free through August 16, which has passed, so budget for it: pricing starts at $0.17 per second. A ten-second clip runs under two dollars. Caplan’s warning is worth repeating: “Don’t search for Flux in the app store. There’s no official app, and lots of imitators use Flux in their name.”

Why these belong in this issue. Most of what’s here took a bargaining committee and years of pressure to win. This is the other kind of gain, the kind you just take. A comms shop of one at a small advocacy org can now put out motion graphics that read as professionally made, for the price of a sandwich, without asking anyone’s permission or waiting for a budget line. That’s the same argument as the Number of the Week, in a different register: the tool is worth the most to the people who were handed the least.

Start with one thing you already publish every week and remake it. Don’t start with a brand refresh.

Source: Wonder Tools, 15 August 2026

Looking Ahead

• California’s WARN Act recommendations land around November. They come out of Newsom’s May 21 workforce order on a 180-day clock. Issue 16 covered that order when it was signed and Issue 17 flagged this as the test of whether it has teeth. Still the date I have circled.
• CHAI’s PULSE pilots are expected this fall. The program launched in July to help public health agencies put AI to work at scale. No jurisdictions have been announced yet. When they are, that’s the deployment story Issue 22 could only cover as a promise.
• SB 947 goes to the Assembly floor. If it passes, California will have written a human review requirement into employment law, and the question stops being whether this idea is viable and becomes which state copies it first. A lot of things got promised in the spring. Fall is when we find out which of them were real.

Until next time,
Jordan

Read past issues  ·  RSS  ·  Website