by: Anna

In 2024, media workers of the Ziff Davis Creators Guild (part of the NewsGuild of New York, local 31003) won pro-worker AI language in their first successor collective bargaining agreement. Despite being a union of 62 workers among a corporate conglomerate of 4,000, we came together to fight the boss and set up AI guardrails. 

Here are the highlights of the AI section of our contract:

• Definition of “generative artificial intelligence”: We specified generative AI as systems that can generate content, not existing technologies that can assist with content creation like spellcheck and Grammarly. We don’t consider those things to be threats to our jobs.

• No layoffs due to the implementation of generative AI! And no reduction in base salary (meaning, no reduction in someone’s annual salary as opposed to a bonus) 

• Implementation of an AI subcommittee where unit members and management will come together and discuss issues that come up including if the company plans on implementing anything that will impact our jobs. Management must give the subcommittee and thus the members reasonable advance notice of this

• Editorial integrity: We have a separate article about editorial integrity that already existed from our first contract but we wanted specifics about AI here. 

• Any AI content must be done at the direction and editorial review of human beings with editing duties — this is very specific because we didn’t want to create a situation where reporters whose jobs are to write would now be saddled with this work

• The company must provide transparency around use of AI on-site for our readers in the form of disclosures both in text and multimedia

How did ZDCG win AI protections?

As always, we won by coming together and taking collective actions. We’ll break this down into several points:

• Digital and in-person actions

• Using management’s own words against them

• Rallying members with current events

• Simplifying multiple messages into one cohesive message

Digital and in-person actions

We—our bargaining committee, our contract action team (CAT) who spoke to members and organized them, and our unit as a whole—used two metaphors to think about collective action this campaign: the campaign mountain and pressure thermometer. The former refers to escalating actions to reach the peak of strength, which is withholding labor through a strike. The latter holds the idea to start with low pressure actions and “raise the temperature” to the most high-pressure action—which, again, is a strike.

For this campaign, we planned that peak of strength around the July 2024 Amazon Prime Days because those are huge moneymakers for the company and withholding our labor would really hit management where it hurts. (We ended up building to a strike threat, but not striking.)

On the pressure thermometer, we started online and worked our way up to in-person. As most of our union is made up of remote workers, we relied heavily on Slack, email, and later Twitter for these actions. 

Slack and email really worked for us. It was an opportunity for a public yet internal action. For Slack, we alternated between spamming channels that included our whole division—so more than our union but not the whole company—to a channel that did basically include everyone, the #general channel.

Twitter also worked later in the campaign as we went up the pressure thermometer—the company hates public scrutiny so we knew we had power there.

As the campaign went on, those who could did meet in-person. We did a one-hour rally partyway through our campaign really as a test-run to an eventual strike. While we ended up finishing the contract ahead of a strike, it was great to meet up for some IRL solidarity and “practice.” And later towards the very end, we did hours of informational picketing outside the company’s office.

Using management’s own words against them

There was a lot of back and forth about AI at the bargaining table (Zoom room), and off of it. But at first, we weren’t seeing much movement from management when it came to the actual proposal for contract language. 

At one early session, a manager said that she didn’t want to use AI for original content and it would be “committing brand suicide.” And in a staff meeting, another manager said if an AI mandate came her way she’d discuss it with us. So the day before the next session, when management owed us a counter AI proposal, we executed a reply-all email action calling out these managers for what they said, and how it was hypocritical compared to their non-movement at the table.

Our union chair kicked off the action, CCing them, their boss, and all our members. A supermajority of members replied to this email calling the EICs out by name and telling them to come to a fair agreement on AI.

At that following session, management presented us with: “The Company will not layoff bargaining unit employees solely as a result of the implementation of Generative AI technology.” 

After back-and-forth we won what we have now; they removed the word “solely” and also added that no one will get a decrease in pay. It was around that time that we homed in on the editorial integrity language as well. 

This was such a victory, both no layoffs and editorial integrity, but the company was being pretty shady about their AI plans. They would give us non-responses to our information requests, and they said at the table there aren’t any plans but then refused to answer any questions. And in terms of our proposal, they refused to give us notice on AI implementation and notice on training AI on our content. And they also struck the idea of an AI subcommittee and said we can discuss these issues in our typical labor management committee meetings—so there was definitely more work to be done.

Rallying members with current events

In the months we were bargaining, partnerships between OpenAI and publishers like Vox and NewsGuild sibling The Atlantic became public which riled up our membership and really drove home that we needed notice if something similar were to happen. Notice doesn’t mean that deals are stopped, but we can at least organize and make the company have a very bad time (and perhaps rethink).  

While some members didn’t care about AI at the start of the bargaining cycle, instead wanting to focus on wages and healthcare (understandably), AI became more and more of an issue in the cultural zeitgeist and media industry on the whole, and thus to our members. The bargaining committee and contract action team used current events to bring members together on the issue.

Simplifying multiple issues into one message

The AI sticking points as well as unrelated issues (such as management trying to withhold backpay, which is another story) made us emphasize the lack of trust between us and management.

Close to the end of the campaign, we kicked the Slack action up a notch with what we called Spartacus Slack. Named after the famous movie scene where everyone claims to be Spartacus, we all changed our Slack profile pictures to that image to see on the right and we all changed our display names to “Fair Contract Now” for a week so to figure out who anyone was you had to hover over their names. 

And we also dropped messages in another channel, this one just of our division at Ziff Davis, so more than our publications but less than the entire company. The message was a poem one of our CAT members wrote: “How can we trust management, if they offer back pay and pull it away? / How can we trust management, if they train AI without our say? / How can we trust management, if they push remote protections away? / How can we trust management, if they won’t give us competitive pay?”

We had our in-person lunch rally afterwards with similar messaging: trust us, trust the workers. This was also reflected in our strike pledge which is a little hard to read now but really drove that message as well. 92 percent of the unit signed it. Even though we had multiple issues going on, we found that throughline of trust that brought it all together. 

At the last bargaining session management agreed on that point of reasonable advance notice which was wonderful. While we didn’t win every single point we wanted in our AI proposal, we won a lot, and it was because of collective action.