Remove yourself from AI training data

Every major AI vendor built their first model by scraping the open web. Your name, your bio, your old blog posts, your social profiles, your data-broker listings — all of it likely ended up in a training corpus along the way. Now those models answer questions about you, surfacing and synthesizing what they found.

What removing yourself from AI actually means

There are two distinct things to opt out of, and most vendor pages conflate them.

Future training. Stop the vendor from using new data they collect about you — your future conversations, public posts they re-scrape — to train new models. Almost every major vendor offers some version of this opt-out. Effectiveness varies.

Past training. Remove information about you that's already been incorporated into a trained model. This is harder. The data isn't sitting in a database waiting to be deleted — it's been compressed into model weights. The realistic ask is: don't let the model surface specific facts about you in responses. Vendors enforce this through output filters and fine-tuning, not literal deletion.

The 2023–2025 lawsuits — NYT v. OpenAI, Authors Guild, and others — are forcing this distinction into the open. Several vendors have introduced "right to delete" flows in response. The flows are real but vary widely by vendor.

What Delist files for you

Delist files the vendor-specific opt-out and data-deletion requests on your behalf. For the major US vendors, that means:

• Filing the per-vendor "do not train on my data" request through each vendor's documented flow.
• Filing the per-vendor data-deletion request where one exists.
• Watching for new vendor opt-out flows as they're published — the AI landscape moves fast, and new models keep appearing.
• Re-filing when vendors update their terms or a request expires.

What we don't claim: we can't make a model forget you. Nobody can — not us, not the vendor, not a court order. What we can do is file every documented opt-out, file them on a schedule, and document the requests in case enforcement opens up.

The supply-chain problem

Filing the vendor opt-outs handles the legal and account-level layer. There's a second layer those opt-outs don't reach: the public data brokers that pre-training data flows through. If your name and address are still on the people-search sites, anyone — including the next AI training run — can re-scrape them.

The full defense runs in both directions: opt out of the AI vendors directly, and remove the source data from the broker layer the vendors scrape. Delist covers both.

What filing opt-outs doesn't do

Honest about the limits:

• It doesn't unlearn you from existing models. The data is in the weights.
• It doesn't reach AI models that were trained on web crawls before vendor opt-outs existed.
• It doesn't apply to foreign or research models that don't honor US opt-out requests.
• It doesn't reach models trained on data brokers' licensed data — that's the broker-layer problem the broker opt-outs address separately.

What it does: file every documented opt-out on a schedule, file deletion requests where the legal basis exists, and cut future training-data exposure to the floor of what's currently achievable.