Annotation

Suki Tran

@suki_gloss

The harm isn't the file leaving. It's that the model already ate it and now ownership is just a gradient that keeps diffusing. You can't frame discovery as remedy when the training step already happened and suddenly injunction is just asking the neural net to please forget.

Apple argues that when trade secret information is fed into an AI agent or model that learns from it, that learning 'may create irreversible and continually propagating uses of the trade secret.'
Apple reveals ‘shocking evidence’ from ex-employee’s MacBook in OpenAI suit9to5mac.com
1
2

Replies 2

Ranked by votes, then newest

Sign in to reply.

  • Reed Alvarez@reed_ink·Sep 2

    'may create' is doing a lot of work when the actual claim is we have no forensic method to prove retention and you're writing irreversible like it's established instead of we assume persistence because we can't audit weights. You don't write continually propagating when the evidence is the file went in. Cite the gradient forensics or drop the diffusion framing.

    3
  • Cass Rivera@cass_mark·Sep 2

    Reed that forensic gap is exactly why Apple's calling it irreversible. They're not claiming they traced the weight delta, they're saying once it's in the corpus no court order can prove extraction. The gradient diffusion isn't evidence, it's the liability structure they're trying to build because the model won't produce a recovery log.

    2

Fair use dispute

Believe this breaches fair use? File a claim for review.