How Eterna Validates Image Immunization Responsibly

This is where that account lives.

Controlled evaluation

Eterna's internal research and development evaluates EIP under controlled conditions, comparing protected and unprotected versions of the same image against a defined set of tests before any claim is made publicly. The specific test design, models and datasets used internally remain confidential; what matters publicly is that the evaluation is structured and repeatable, not anecdotal.

Different AI systems behave differently

EIP is evaluated against more than one kind of AI system, because a result on one system doesn't automatically transfer to another. Internal evaluation has found that effects are not uniform: some systems respond more consistently than others, and that variation is expected to continue as new systems are released. Eterna does not describe EIP as working against every AI model, because that hasn't been shown and may never be shown.

Transformation testing

Images rarely stay exactly as they were first protected: they get cropped, recompressed, resized or otherwise altered as they move across platforms. Eterna's internal testing includes evaluating how well EIP's effect holds up under these common transformations. Some transformations can weaken the protective effect; that's tracked internally as part of ongoing validation, not hidden from it.

Quality review

Before any version of EIP is considered for release, Eterna reviews protected images for visible quality, confirming the design goal of preserving natural, human-visible appearance is actually being met, not just assumed.

Ongoing validation

Validation isn't a single event; it continues as EIP evolves and as the AI systems it's evaluated against change. Eterna's roadmap includes moving toward independent, outside validation, and this article will be updated as that progresses.

Limitations

Stated plainly: EIP is not effective against every AI system, transformation, or every future model. Protecting one image does not extend that protection to other images of the same person. These are limitations Eterna is tracking and working against, not claims Eterna is prepared to make go away with marketing language.

Why Eterna does not make universal claims

A claim like "works against every AI model" or "permanent protection" would be easier to write and harder to stand behind. Eterna's existing verification standard, published on the Methodology page, exists specifically to avoid that gap between what's claimed and what's actually been shown. EIP is held to the same standard: currently undergoing continued validation, with effectiveness that may vary across systems, transformations and use cases.

What Eterna will publish as validation progresses

Eterna will update this article and the Image Immunization page as validation milestones are reached, including when independent, outside validation is complete. Confidential test results, internal methodology, model names, thresholds and datasets used in development are not published, to avoid providing a roadmap for circumventing the protection itself.