“All past performances are not eligible to be part of training data. Talent will have the option to allow their performance/work to be added to a training data set no less than N years after the first public release”

And how can that ever be truly ensured/compliance confirmed?

Studios can pinky swear then somehow, oops, some third party that they totally didn’t know about their practices uses the off limits material to generate AI results which are then sold to the studio.

It’s unfortunate that these are the sticking points for the union, because they seem like really bad hills to die on long term.

Residuals are antiquated in the streaming world, gettingpaid up front is a far better deal. If you want recurring income, negotiate a stock grant. Residuals are chasing a shrinking pile of money competing with 40 years of content. Prestige TV makes headlines, but a big reason streamers aren’t releasing numbers is because those shows are crap for engagement. Shows like Friends are still likely dominating the streaming count.

AI is coming whether you like it or not, they would be far better off negotiating a way to include AI as it advances rather than seek to ban it. The fair use argument is really strong for model builders. Actors and writers are better off working with current producers to create a path forward for AI, because a complete ban is likely to result in an outside company coming in and disrupting the market. Yes the stance producers started with is extreme on this subject matter, but the actors and writers are similarly extreme.