AFP via Getty Images
Inside the working group talks that formed the guild’s proposals to tame the AI threat
Last summer, Van Robichaux ran for the board of directors of the Writers Guild of America West. Out of 17 candidates, he was the only one who raised a concern about artificial intelligence in his campaign statement.
“As far as I know, this issue is not on the radar of anyone else running for the board and while I might sound like a paranoid lunatic talking about it today, in 10 years I’m confident you’ll be glad I brought it up now,” he wrote.
He did not win.
AI has since become the hottest topic in the creative economy, spurred by the release of models like Stable Diffusion last August and ChatGPT in November. Across disciplines – graphic design, animation, acting, music, writing – artists are terrified of being replaced by robots.
“I think they’re right to be concerned,” said Bruce Schneier, a lecturer in cybersecurity at the Harvard Kennedy School of Government. “Things are changing so fast, that things that were true three months ago aren’t true today.”
AI has also become a central issue in the writers strike, displacing more familiar fears like the rise of streaming and the decline of residuals. Writers on the picket lines fear that movie studios will use AI to write scripts – either in whole or in part – diminishing the role of writers or even making the job obsolete.
“The corporations will push us all into extinction if they can,” said Chap Taylor, a screenwriter and professor at New York University. The AI issue “is life and death,” he said. “That’s the one that turns us into the makers of buggy whips.”
Though he did not win his election, Robichaux was partly responsible for pushing AI into the center of the conversation. After his defeat, he was approached and invited to join the guild’s AI working group. Many of the writers in the group worked on sci-fi shows and had a deep interest in technology and computer science.
Over several months of Zoom meetings, they crafted what would become the WGA proposals on artificial intelligence – the framework of regulations that would be presented to the studios during collective bargaining.
Those proposals began with an appreciation for both the threat and possibilities of AI. They did not seek to outlaw it. Instead, they focused on protecting screenwriters from economic harm, while allowing them the freedom to use AI if they want to. The framework was fundamentally optimistic, holding that AI could be both constrained and harnessed for good.
“I don’t think you can ban a tool,” said John Rogers, another member of the group. “What we’re saying is, ‘Use it as a tool.’”
What they were trying to do was ambitious. They were trying to coax multibillion-dollar corporations to accept their vision of the future, and to swear off competing visions. Their only leverage was the threat of a strike.
Getting to yes would also take more than just leverage. It would take a healthy amount of communication, sophistication, and trust. But after weeks of talks in March and April, the AI conversation dissolved in mutual suspicion and misunderstanding.
When they called a strike, WGA leaders said that the studios refused to rule out using AI in the future. The Alliance of Motion Picture and Television Producers, the organization that bargains on the studios’ behalf, has not denied that.
But in their proposal, the writers aren’t ruling it out either.
At this stage, it’s not clear whether AI will be helpful or legally permissible in screenwriting. But if it passes those tests, it’s hard to imagine it won’t be used. The real question may be, “On whose terms?”
The AI working group was made up of seven or eight people and led by Deric Hughes, a WGA West board member who has written on shows such as “Quantum Leap,” “Arrow,” and “The Flash.” Another member of the group, John Lopez, also writes on sci-fi shows, and among his specialties is dreaming up the mysterious technologies wielded by alien civilizations.
They talked on email chains and traded text messages about the latest developments. They also experimented extensively with AI models and began gaming out scenarios. One thing they determined early on was that in its current form, AI isn’t good at writing screenplays.
“It’s good at summarizing things,” said Robichaux, whose credits include “Brooklyn Nine-Nine.” “It’s good at trying to kind of complete lists of things. It’s not so great at the kind of work we do.”
“The capabilities are wildly overblown,” he said. “A lot of this hype is because Silicon Valley needs the next big thing and they don’t have one. So this is it.”
Lopez spent hours on ChatGPT, trying to get a better grasp of the threat. He fed it prompts over and over, trying to generate something worthwhile.
“It took almost as much work as writing it from scratch myself,” he said. “It did make me freak out a little bit less.”
Aside from functional problems, there are also legal hurdles. AI models train on massive troves of published material, much of which is protected by copyright. Lopez said a friend referred to AI systems as “plagiarism machines.”
The developers of these models have argued that the output is “fair use.” But many in the creative world are starting to raise alarms, and file lawsuits, about the misuse of protected material.
In part for that reason, Robichaux said he doubts any studio would bother trying to use AI to replace writers during the strike.
“I have worked with all of the studios in town,” he said. “There is no chance the legal department is letting them actually use any of these AI models that are trained on other stuff… These people won’t let you wear a Nike shirt. I don’t buy for a second that there’s any sort of meaningful computer scabbing going on. I just don’t buy that threat.”
Some members of the group were surprised, in fact, that the studios have not already filed lawsuits to protect their own IP from being used to train AI systems.
The group had mixed views about whether the technology will ever be good enough to replace writers.
Lopez worried it could become more of a threat over time, but he is not quite convinced that it will make writers obsolete.
“I ultimately don’t think there will be a shortcut to the creative process,” he said. “You’re still going to have to figure out what you want to say.”
Given the uncertainty, though, Rogers argued it’s best to put guardrails in place now.
“Should you be worried? Absolutely,” he said. “They’re going to try to use AI to replace writers. Will the shows be good? No. Will that stop them? No.”
Still, the group did not want to ban AI outright. They also did not want to enact a moratorium, which might give everyone breathing space to figure out how to approach the subject.
“Banning it puts us out of the game,” Rogers said. “There’s a lot of hot, dumb money behind this. It’s better to channel it than get run over by it.”
They feared that if they proposed a ban, they would be caricatured as anti-technology. And they did not trust the studios to adhere to a moratorium. Instead, they decided it would be better to assert their interests now, rather than trying to prevent an inevitable future.
“It feels like that was a road to extinction,” Lopez said. “Silicon Valley is going to spend a lot on this. It felt like saying, ‘Let’s ban the internet’ when AOL was starting up.”
In Lopez’s words, the goal was to make sure writers are “not getting crushed by the wave, but figuring out how to surf it and ride it.”
Some of them could also foresee a time when AI is actually useful, in limited ways, to screenwriters.
Lopez said it might be good for condensing a long chunk of text. Robichaux likened it to tools that already exist in Final Draft, which can fill in a character’s name when the writer types the first letter.
“An AI system would be able to, when I hit enter, without hitting ‘C’ it would type ‘Charlie,’” he said. “Now I’m being more productive. I’m using AI, but not in the way where it’s generating the idea for me. There are ways to create new useful tools for writing with this technology that aren’t undermining our process.”
The cases they imagined are similar to the way visual effects artists already use AI — as a tool, but with the creative person still firmly in charge.
Lopez said that for him, the jury is still out as to how useful AI could turn out to be. He compared it to the possibilities created by splitting the atom.
“It’s powerful and dangerous,” he said. “There could be incredible good and incredible harm.”
The AI working group ultimately developed three basic proposals, Robichaux said.
First, AI-generated material would not be considered “literary material” or “source material” under the union’s contract. That would prevent studios from paying writers less, or depriving them of credit, if they rely on AI material.
Second, they said that AI should not be allowed to write on its own. Studios would be forbidden from having AI programs create scripts independently, or having them rewrite scripts submitted by a human writer.
And third, a studio’s AI program would be barred from training on WGA members’ work. If the studios rejected that, guild members might agree to allow it in exchange for a license fee.
Heading into negotiations, the top WGA leaders did not see AI as one of their main issues. Instead, they planned to focus on TV staffing and other core economic concerns. The only public hint that AI would even be mentioned in the talks came on Feb. 27, when the guild revealed a long list of agenda items, including a proposal to “regulate” the use of AI technology – not “ban” it.
The regulatory approach was consistent with how negotiating committee members were thinking about the issue. “Scriptnotes” co-host John August, who is on the committee, was quoted in The Hollywood Reporter in January saying that AI was all but inevitable.
“There certainly is no putting the genie back in the bottle,” he said. “It’s going to be here, and we need to be thinking about how to use it in ways that advance art and don’t limit us.”
August was among the writers who had taken an early interest in AI. He had even invested in Sudowrite, an AI fiction-writing platform. (In a blog post on Friday, August explained how he got involved with the company, and said that he has not made “a cent” from the investment. His picture and endorsement were also removed from the Sudowrite website.)
The AI group presented its working paper to the WGA negotiating committee. Chris Keyser, co-chair of the committee, said in an interview that it received input from multiple sources as it formulated the proposals that would be submitted to the AMPTP. He also said that the working group “was not an official organ of the negotiating committee.”
The proposal submitted to the AMPTP consisted of only one item, according to a studio executive who requested anonymity to speak candidly about negotiations. It was the first item developed by the working group: that AI-generated material would not be considered literary or source material.
It was definitely not a ban. In fact, it contained no limit at all on how AI could be used.
That version – a single sentence – leaked to Variety, which reported that it appeared the WGA wanted to allow screenwriters to use AI as a tool, provided they didn’t lose money or credit because of it.
The story generated both disbelief and outrage. At a time when creative professionals were waking up to the threat of AI, the WGA appeared to be fine with it, so long as it didn’t affect members’ income.
“The Guild doesn’t fear AI as much as it fears not getting paid,” screenwriter Paul Schrader wrote on Facebook. “This I think is the WGA position: If a WGA member employs AI, he/she should be paid as a writer. If a producer uses AI to create a script, they must find a WGA writer to pay.”
As backlash mounted, the guild rolled out a new AI policy on Twitter. The tweets said that the guild’s position was that AI could not be “used” as source material, nor could it be used to “generate” literary material. The tweets also stated that AI material “has no role in guild-covered work.” That left the impression that the guild was trying to ban it.
The topic was not discussed again in the negotiating room until weeks later. At the beginning of April, the negotiators took a two-week break while the WGA obtained a strike authorization vote. When they returned the week of April 17, the writers brought a revised proposal, the executive said.
Despite the tweets, it was still not a ban.
The new proposal stated that AI-generated material “shall not be considered source material or literary material on any project covered by this or any prior Basic Agreement.” In other words, AI material could be used — it just would not count against writers in determining credit and pay. The WGA also proposed that the studios could not train their AI programs on guild members’ work.
Negotiating committee members took turns discussing other issues in the room that week. But when it came time to present on AI, the task was given to Charles Slocum, a longtime member of the WGA staff. Without going into details of the proposal, Slocum warned about the uncertain copyright landscape surrounding AI, and seemed to suggest that AI was too risky to use at all.
Toward the end of his presentation, he invoked “Mrs. Davis,” a Peacock show that pits a nun against an all-powerful AI system. “‘Mrs. Davis’ was written by humans,” he concluded.
The studio lawyers did not know quite what to make of that. The tweets said one thing. The proposals said something else. And instead of trying to clarify what writers want, Slocum was talking about copyright, which is primarily a studio concern.
“They didn’t want to admit out loud that they want to use it,” the studio executive said.
Keyser did not dispute that account, except to say that the proposal did not change over the course of bargaining. He also said he did not recall the precise language of the proposal. Otherwise, he said he would not discuss confidential negotiations.
“Whatever happened in the negotiating room was subject to a press blackout,” he said. “If you heard that, it was from somebody violating the agreement. I’m not going to do that.”
Slocum declined to comment.
The AMPTP lawyers did not come to the table with their own AI proposal. After Slocum’s speech, there was little in the way of engagement on the issue.
The AMPTP rejected the guild’s proposals, but offered a “side letter” that would underscore existing contract language specifying that a writer must be a person. The AMPTP also offered to hold meetings, at least annually, to discuss advances in AI technology.
According to the WGA, AMPTP president Carol Lombardini also said that the studios do not want to rule out using AI in the future. Keyser and negotiations co-chair David Goodman have said that remark was alarming.
AI “was always on our radar,” Keyser said. “It has risen in importance for us in large part because of the studios’ response to what we thought was going to be a simple agreement to restrict it for our and their benefit. We need to be more worried than we thought.”
Appearing before a crowd of 1,800 writers at the Shrine Auditorium on May 3, Keyser warned of a future in which hundreds of shows could work with one writer and a machine. That fear has struck a chord on the picket lines.
“AI has become my number one issue,” said TV writer Chris Duffy, who was marching outside Disney headquarters in Burbank. “I think it’s an existential one. The fact that they refused to negotiate made me be like, ‘Oh, you really want to use it.’”
Kelly Wheeler, also a TV writer, said she too is “most scared about AI.”
“I love writing and I love being around writers,” she said. “And the idea that that creative energy can just be stripped away from television, and instead have a robot do our job – or attempt to – is terrifying.”
From being an afterthought, AI has become perhaps the strike’s most potent issue.
“The more this goes on, the more this becomes a strike over AI issues,” said Michael Colton, co-creator and executive producer of the ABC sitcom “Home Economics. “I don’t think people are feeling like tomorrow AI is going to write a perfect sitcom script. But the fear is that studios will use AI to turn out a crappy first draft, and then turn it over to writers who they hire for a few days or a week to turn it into something good. And they won’t pay them as if it’s an original script. That is the fear.”
Howard Rodman, a past president of the WGA West, recognizes the importance of the guild addressing this amorphous issue.
“When you are out on strike, it’s all about the hopes of the membership and the fears of the membership,” said Howard Rodman, a past president of WGA West. “This certainly speaks to the fears.”
In the interview, however, Keyser acknowledged that the committee is not trying to “ban” AI.
“There are all kinds of uses that AI might be put to,” he said. “Our job is to protect writers from the ways in which AI will undermine the writing process… That’s what we’ve done and we stand by that.”
Robichaux put it much the same way.
“It sometimes get reported as, ‘The WGA is asking studios to not use AI,’” he said. “That’s not the current ask. The current ask is that you cannot use AI to undermine our contract and stop us from getting protections that would otherwise be guaranteed.”
The members of the guild’s AI working group were not in the negotiating room, and were not privy to what happened there.
But they were dissatisfied with the AMPTP’s response. Some interpreted the offer of annual meetings as a stalling tactic that would allow the studios to get a head start on using the technology to undermine writers’ economic standing.
“Whenever a company says, ‘Absolutely not, we’re not discussing it,’ that means there’s a roomful of humans trying to figure out how to do it,” Rogers said. “The AI line is a hard line for us. At some point they will have to come up with a counter.”
Robichaux had not heard about the side letter. But when told of it, he said it would not be sufficient to allay writers’ fears.
“The concern would be that this is a way for them to say one thing, but leave themselves the wiggle room to do another thing,” he said.
The WGA proposal would give writers a lot of flexibility to use AI as a tool. It contains no language limiting the amount of AI material a writer could incorporate. Any limits would have to come from copyright law, which is unsettled on the subject. In March, the U.S. Copyright Office issued policy guidance, saying that material generated purely by a computer — without any human authorship — is not eligible for copyright protection.
But at the same time, the office stated that if a person selects or arranges AI-generated material, that could still be eligible. A writer could potentially use AI to brainstorm, create lists of character names, come up with plots, or even churn out rough drafts, and still qualify for copyright protection so long as the writer remained in control of the finished product.
Hollywood writers must sign certificates of authorship, in which they represent that the work they are submitting is original and copyrightable. If AI becomes a significant factor in screenwriting, it might be important to require disclosure of how, or how much, AI material was used.
The studios have been monitoring the copyright issues around AI for years, especially with regard to visual effects and post-production. The Motion Picture Association, which lobbies on their behalf, said at a U.S. Copyright Office hearing last week that they see “great promise in AI.”
“The studios see it as a tool,” said Ben Sheffner, associate general counsel of the MPA, in an interview. “They do not view it as a way to replace humans. It’s a tool, and it’s going to make the filmmaking process better and make better experiences for audiences.”
Sheffner made clear in his testimony that the studios are wary of being hemmed in by AI regulation. He argued that courts and regulators should “not jump to early, definitive conclusions” based on limited experience.
The studios may be just as reluctant to accept constraints from the WGA, which could be hard to change in future rounds of bargaining. Despite that, Robichaux said he expected that the WGA and the companies can come to a reasonable agreement.
“It’d be great if the studios would say, ‘Hey, we do need rules,’” Lopez said. “Regulations exist for a reason. And it’s a hard job. It requires discussion, consensus and consent. The guild wants to do that. The AMPTP is not an active participant in that offer.”
Jennifer Maas and Cynthia Littleton contributed to this story.