The Writers Guild of America has a deal! Shows will come back! But would-be AI-overlords are still at work.

When ChatGPT became widely available in November 22, AI went from science fiction to science fact. It was a thunderbolt felt around the world, where people from all walks of life became focused on the new tech. Toy, tool, threat — what are we supposed to make of this? AI software and hardware is being applied to all kinds of tasks, from writing to pictures to medical diagnosis, from weather forecasting to love advice. 

Among the fake Shakespeare and distorted portraits, there were actually some really interesting things generated. But most of the attention was focused on flashy, low quality but potentially acceptable imitations of famous and successful work. People interested in exploiting the technology began to mass produce novels and images that may not have been the best quality, but potentially could sell.

This added a whole new category of concerns for the Writers Guild of America (WGA) that was already considering a strike over other long simmering issues. When negotiations began in April the producers refused to even discuss AI in the contract. Now, there is a precedent-setting deal.

The short version: the new WGA contract states that AI can’t be used to write or rewrite any scripts or treatments, requires the studios disclose if any material given to writers is AI-generated, and protects writers from having their scripts used to train AI without their agreeing¹. Provisions in the contract also stipulate scriptwriters can use AI for themselves if they wish. At a time when people in many professions are rightly concerned that they will be replaced by AI, this is a big breakthrough. The writers strike worked!

Here’s a small excerpt from the contract terms that shows how carefully it’s worded:

“Employment Conditions Regarding GAI:

– A company cannot make the usage of GAI that generates literary material a condition for employing a writer.

– However, companies can require writers to use GAI for other tasks, such as detecting potential copyright infringement.”

Here is the official summary from the WGA.

While this deal gives writers some leverage with major studios and streamers, it might not affect an AI company, which may or not be based in the US, from scraping their work. John August of the writer’s negotiating committee concurs, saying the WGA needs “to be honest” about the limitations of the contract. “We made a deal with our employers, the studios,” he says. “We have no contractual relationship with the major AI companies. So this is not the end of the fight.”

In fact, on the day the writers strike deal was announced, the streamers announced the formation of a new Streaming Alliance (SIA)². If this sounds vaguely ominous, consider that their purpose is to fight “regulatory threats.” This new alliance is being set into motion to try to stop government from interfering with streaming services doing whatever they feel they want to do, and could have serious ramifications in the fight against the misuse of AI.

The WGA has achieved an historic deal, a great achievement for professional writers and also a precedent for others. It’s clear the streaming services don’t like this and are trying to find a way around it now.

Simon Johnson, an economist at MIT, says that the WGA has taken the lead in unions addressing AI, and everyone should take note. As he and several coauthors laid out in a recent article on pro-worker AI, the history of automation demonstrates that workers cannot wait until management deploys technology; if they do, they will be replaced. Power and Progress: Our 1,000-Year Struggle Over Technology and Prosperity is his new book on tech.

In the full WGA contract with the studios and streamers (AMPTP), there is an entire section defining different types of AI and their use. Traditional AI is now accepted as part of the workflow in visual effects, and operations management. Generative AI is defined as being able to generate content. 

There are many different kinds of content, however, and it’s clear that enforcing the guardrails around using GAI will be challenging. 

For example: An author uses AI to generate a book, but doesn’t tell anyone. The studio options the book and gives it to a writer to adapt. It’s a clear violation of the contract but how could you prove it? For example, OpenAI, the creators of Chat-GPT, have themselves have admitted that there is no reliable way to detect when AI has been used in this way. They can’t detect the use of their own tool. What if the studio uses data gathered from streaming viewing habits to train AI, which in turn is used to generate notes on how to turn a writer’s screenplay into something more like what their audience likes? The human would literally be working under the supervision of an AI-overlord, and not know it.

It’s been suggested that all software makers and AI hardware makers embed watermarks that will show if content was made using AI. While that would require legislation, and might be probably the only practical way to enforce the terms of the writers contract, in the case of text generation there may be no way to effectively do that. Until this question is resolved, we have to trust the studios will do the right thing.

And that concerns the actors Guild, who are still on strike. Until both are resolved many shows will remain struck and in a suspended state of production. The Writers Guild of America honors other Guilds strikes in solidarity.

There is cause to celebrate, this is a major step forward. More news to come.

-30-

¹ This is an important detail, since copyright law does not currently address the issue of copyrighted work being used to train AI. Companies that produce artificial intelligence products are relying on the legal principle that it is not illegal to read and interpret written material to produce analyses or to inform one’s own work regardless of who (or what) does it.

2 A new trade group — the Streaming Innovation Alliance — brings together Netflix, Disney, Warner Bros. Discovery, Paramount and other players to promote their interests to politicians and government entities. It marks the first time competing streaming-video providers have banded together in a unified lobbying front, and represents a powerful political force .