A TV writer named Sarah spent 15 years getting credited on network shows and a streaming limited series. Now she rates AI-generated screenplays for roughly $25 an hour on annotation platforms like Scale AI's Outlier.
"I'm literally teaching it to replace me," she told Wired's Ruth Fowler in a
4,900-word investigation published Sunday, "and I need the money too badly to stop."
Fowler spoke with 47 former entertainment professionals doing the same thing: annotating, correcting, and refining AI outputs for companies like Scale AI, Surge AI, and DataAnnotation.tech. Many estimated 30 to 50 percent of their professional circles are doing identical work. All under sweeping NDAs. The work is eerily specific to their expertise. Writers evaluate AI scripts. Editors do frame-by-frame analysis of AI footage. A former showrunner evaluates AI-generated series bibles for $45 an hour.
The pay tells you what the market thinks their knowledge is worth when they don't set the price.
Then 400 of the most famous people in entertainment set the price.George Clooney. Tom Hanks. Meryl Streep. Scarlett Johansson. Leonardo DiCaprio. Over 400 performers backing RSL Media's Human Consent Standard, a framework co-founded by Cate Blanchett and CAA co-chairman Bryan Lourd.
The framework introduces a universal "consent key" (a machine-readable signal performers can set to allow or deny AI use of their likeness) and a Digital Identity Registry launching in June 2026. Performers set terms, grant permissions, and track every usage request. SAG-AFTRA endorsed it. Studios are in talks to adopt it.
"Our likeness is not up for debate," the open letter reads. "It belongs to us."
Same industry. Same week. Completely different leverage.The difference between Sarah rating scripts for $25/hour and Cate Blanchett building a consent registry isn't talent. Sarah has 15 years of network credits. The difference is that Blanchett decided what her expertise was worth before someone else decided for her.
That's the pattern worth paying attention to, regardless of your industry.
The numbers frame the scale.Bureau of Labor Statistics data shows employment in motion picture and sound recording fell 26 percent between 2022 and 2025. A CVL Economics report estimated 204,000 entertainment jobs affected by AI. The global data annotation market hit $5.8 billion in 2025 and is projected to reach $22.1 billion by 2030.
Scale AI, the company behind Outlier and Remotasks, is valued at $13.8 billion. Its contractors have no equity, no benefits, and no job security. Marcus, a former reality TV editor, earns about $2,200 a month through annotation. He used to make $5,800.
An informal collective has started meeting in Los Angeles. They call themselves the Ghost Workers. After Fowler's article, an emergency meeting drew 60 people. Their Signal group hit 200 members in two days. Their demands: transparent pay, client disclosure, residual structures, benefits access, and protection from retaliation. They're organizing. That matters.
The uncomfortable math is simple.The annotation workers have the same specialized knowledge as the A-listers. The difference is structural: who set the terms before the work started. The WGA acknowledged their current contract doesn't address the "reverse scenario: writers being used in the AI training process." They plan to make it central when the MBA expires in May 2028. State Senator Scott Wiener announced legislation requiring AI companies to disclose their use of entertainment professionals and guarantee minimum compensation.
Sarah said the AI scripts she evaluates have gone from laughable to competent in six months. "Not good. But competent. And competent might be enough."
She gives it another year. Maybe two.
