Some artists who tried to use Meta's request form to delete their data from its AI training say the new system is broken; one calls the form a “fake PR stunt”
This summer, Meta began taking requests to delete data from its AI training. Artists say this new system is broken and fake.
Context & Ripple Effects
Meta had only recently introduced a channel for people to ask to access, alter, or delete third-party information used in generative-AI training, making artists’ accounts an early test of whether that new data-rights process could work in practice.
The dispute matters because later criticism of image-generation features tied to public Instagram accounts led Meta to discontinue the capability, underscoring how consent and redress mechanisms can become product constraints rather than back-office policy.
First-order effects
- Artists seeking removal from Meta’s AI-training inputs may be unable to obtain a usable response through the announced process, leaving their data-status requests unresolved.
- Meta’s claimed opt-out and deletion pathway faces an immediate credibility problem among the creators it is meant to address.
Second-order effects
- A process perceived as ineffective gives creator advocates a concrete operational failure to challenge, rather than a purely abstract objection to AI training.
- Product teams building generative-AI features on public or third-party data may face greater pressure to make provenance, removal, and escalation workflows auditable.
Third-order effects
- If opt-out systems cannot reliably translate requests into training-data controls, the industry may shift toward more explicit permission boundaries for public data rather than relying on post hoc request forms.
- The later withdrawal of an image feature involving public Instagram accounts suggests that disputes over likeness and training-data control can increasingly shape which generative-AI features remain deployable.
The trend: Generative-AI platforms are moving from broad public-data collection toward a contested governance layer in which consent, deletion, and likeness controls determine product viability.