InverityYour do-not-train preference, carried in the file.
Aegis embeds a machine-readable do-not-train signal in your images and emits the matching web signals, so AI crawlers get a clear, standards-based answer. It reads fail-closed, so silence never gets read as yes.
You never agreed to be training data. Saying so is the hard part.
Opting out of AI training today means scattered, ignorable half-measures: a line in a terms page nobody parses, a setting on one platform that does not travel, a preference that vanishes the moment your image is downloaded or reposted. There is no single, portable way to say no that a crawler will actually read.
And the default cuts against you. When the signal is missing or unclear, most scrapers treat that as permission. What you need is a preference that is machine-readable, standards-based, travels with the file, and is safe by default.
Set your preference once. It travels wherever the image goes.
Aegis runs on the same connector layer as the Neural Media Orchestrator. As assets flow through your pipeline, Aegis writes a do-not-train preference into the file using IPTC and PLUS Data Mining fields and XMP rights metadata, and emits the matching web signals: robots.txt AI crawler directives, the W3C TDMRep reservation, and AIPREF preference signals. It also works with do-not-train registries.
The signal reads fail-closed. If it is missing or ambiguous, the safe reading is do not train. Aegis is one layer of the Provenance suite: it declares consent, Attest proves origin, and Signet proves ownership.
One preference, every place a crawler looks
In the file
IPTC and PLUS Data Mining fields plus XMP rights, embedded so the preference travels with the asset, not just the page it sat on.
On the web
robots.txt AI crawler directives, the W3C TDMRep reservation, and AIPREF preference signals, so the answer is present wherever a crawler checks.
Fail-closed by default
If the signal is missing or ambiguous, the safe reading is do not train. Silence is never treated as consent.
Registry-aware
Works with do-not-train registries so your preference is discoverable beyond the file itself.
Works across your connectors
Apply the signal as assets move through your DAM, CMS, or store, on the same connector layer the orchestrator already uses.
Part of the Provenance suite
Stacks with Attest (signed provenance) and Signet (forensic watermark) for consent, origin, and ownership in one line.
By the standards
- IPTC and PLUS Data Mining fields, plus XMP rights metadata
- W3C TDMRep reservation and AIPREF preference signals
- robots.txt AI crawler directives and do-not-train registries
Safe by default, honest about the limits
Aegis reads fail-closed, so the safe reading is always do not train. We are also straight about what a signal is: it is a clear, standards-based statement of your preference, not a technical block. It is honored where crawlers respect these standards.
For proof that survives regardless of who honors the signal, Aegis pairs with Attest and Signet.
Questions worth answering straight.
What does a do-not-train signal actually do?
It states your preference in a machine-readable, standards-based way, both inside the file (IPTC, PLUS, XMP) and on the web (robots.txt, TDMRep, AIPREF), so a crawler gets a clear answer instead of a guess.
Will AI companies honor it?
It is honored where crawlers respect these standards. A signal is a preference, not enforcement, so we are honest that it is not a technical block. That is exactly why it pairs with Signet, the forensic watermark, and the training-detection research, which do not depend on anyone’s goodwill.
What does fail-closed mean?
If the signal is missing or ambiguous, the safe reading is do not train. Silence is never treated as permission.
Is this the same as a watermark?
No. Aegis is a preference signal that travels with and around the file. Signet is an imperceptible mark in the pixels that proves ownership. They solve different problems and work best together.
Which platforms support it?
Aegis writes signals through the Inverity connectors. Tell us your stack, a DAM, a CMS, or a store, and we will confirm the exact path for your setup.
Say no, in a way that travels
Tell us your stack and we will show you how Aegis sets your preference across it.