Legal

Privacy policy

Last updated: September 22, 2026

1. Who is responsible for your information

Legacy Studios, Inc., a Florida not-for-profit corporation doing business as Eternal Library Foundation, with its principal office at

PRINCIPAL OFFICE ADDRESS

is responsible for the personal information described in this Policy.

Where United Kingdom or European Union data protection law applies to our processing, Legacy Studios, Inc., doing business as Eternal Library Foundation, acts as the controller of the personal information described in this Policy.

Questions, and requests concerning your information, should be sent to privacy@eternallibrary.ai.

2. What this Policy covers, and who processes information for us

This Policy covers personal information controlled by Legacy Studios, Inc., doing business as Eternal Library Foundation. It does not govern processing performed by independent third parties for their own purposes.

Materials deposited with us under a written deposit agreement are also governed by that agreement and remain subject to applicable law.

Three layers, one standard

The service you use rests on three layers. Each has its own commitments, and they are held to a single standard.

Eternal Library — the layer you deal with. We decide what is collected, why it is collected, and how long it is kept. The commitments in this Policy are ours. They are the floor beneath everything below.

Archivista — the application layer. Archivista Archives, Inc. provides the platform on which collections are preserved, described and presented, under a written agreement with us, and processes personal information only on our instructions and for no purpose of its own. Archivista's own published commitments are adopted here and passed through to depositors and visitors: that data entrusted to the platform is never used to train external AI models, by any route including anonymization or aggregation; that ownership of deposited content stays with the depositor; that nothing is shared across tenants; and that content can be exported in standard, portable formats.

Microsoft — the infrastructure layer. The platform is hosted on Microsoft Azure and uses Microsoft Foundry for model inference. For that layer we adopt, and pass through to you, Microsoft's contractual data protection commitments: the Microsoft Products and Services Data Protection Addendum, which governs processing by the Azure and Foundry services in use; the Microsoft Foundry data privacy commitments, set out in section 5; and Microsoft's published privacy principles of user control, transparency, security and defending data.

Where a commitment at the Archivista or the Microsoft layer is stronger or more specific than ours, you get the benefit of it. Where ours is stronger, ours governs. Adopting another party's commitments never dilutes the commitments in this Policy, and never moves our accountability to someone else: you deal with Eternal Library, and Eternal Library answers for the whole chain.

3. Information we collect

Information you give us

  • your name, email address, institution and any other detail you choose to include when you write to us or complete a form on the Site;
  • the content of your correspondence with us.

Information collected automatically

  • your IP address, browser type and language, device and operating system, referring page, and the dates and times of your visits;
  • aggregated and de-identified information about how the Site is used.

Information generated when you use an AI tool on the Site

When you use a collection-specific discovery tool or an interactive avatar, the service processes what you ask it and what it answers, in order to answer you. What is captured and how long it is kept is stated in sections 5 and 10.

WHAT IS CAPTURED IN A DISCOVERY-TOOL OR AVATAR SESSION — QUERY TEXT, VOICE, CONVERSATION HISTORY, SESSION IDENTIFIER — AND WHAT IS STORED: TO BE COMPLETED BY DANIELA PEÑA. STATE ONLY WHAT IS ACTUALLY CAPTURED.

Information we do not collect

We do not collect or store payment card numbers, private cryptographic keys or seed phrases.

We set no advertising identifiers and we do not track you across other websites or services.

BIOMETRICS: IF AN AVATAR ACCEPTS VOICE OR IMAGE INPUT, STATE WHETHER ANY BIOMETRIC IDENTIFIER IS DERIVED FROM IT. IF NONE IS, SAY SO PLAINLY. DO NOT PUBLISH A DENIAL THAT HAS NOT BEEN CONFIRMED.

The Site is not directed to children under 13 years of age, and we do not knowingly collect personal information from them. If we learn that we have received personal information from a child under 13, we will delete it.

4. How we use your information

  • to respond to your inquiries and to maintain our correspondence with you;
  • to provide the collection-specific discovery tools and avatars you choose to use, and to return an answer to what you ask;
  • to operate, secure, maintain and improve the Site;
  • to send you information you have asked to receive, and to stop sending it when you ask us to;
  • to keep the records we are required or reasonably expected to keep; and
  • to comply with legal obligations and to protect our rights and those of others.

We do not sell, rent or trade personal information, and we have never done so. We do not use it for advertising. We do not use it for automated decision-making that produces legal or similarly significant effects.

5. AI and collection-specific tools

Eternal Library uses artificial intelligence to power collection-specific experiences, including discovery tools and interactive avatars. This section states what those tools are built on, what we never do with what you entrust to us, and what the infrastructure beneath them commits to.

How a collection-specific tool is built

SELECT THE ONE THAT IS TRUE, AND PUBLISH ONLY THAT ONE — DANIELA PEÑA:

ALTERNATIVE A — RETRIEVAL. "A collection-specific tool answers by retrieving material from the deposited collection it serves and placing that material before a general-purpose model at the moment of the question. The collection stays where we hold it. The underlying model is not altered by it, and holds nothing from it afterwards."

ALTERNATIVE B — FINE-TUNING. "A collection-specific tool is produced by fine-tuning a model on the deposited materials of the single collection it serves, under the terms of that collection's deposit agreement. The resulting model is available exclusively to us and to that depositor, is encrypted at rest, and can be deleted at any time."

Whichever applies, the following is true of both, and is published in either case: no collection is used to alter a model that serves any other collection or any other organization; these models are not open systems; and no deposited material is used to train, retrain or improve a general-purpose or foundation model.

What we never do

  • We never use deposited materials, or your interactions with our tools, to train, fine-tune or improve any general-purpose or foundation model, whether ours or a third party's.
  • We never permit a service provider or a model provider to retain for training, or to learn from, material entrusted to us.
  • We never use anonymized, de-identified or aggregated versions of that material for model training. De-identification claims carry re-identification risk, so we do not rely on them. The exclusion is absolute.
  • We never let one collection inform another. What a tool draws on in one collection serves that collection alone.
  • We never sell or rent personal information, and we never use it for advertising.

The infrastructure these tools run on

Model inference runs on Microsoft Foundry, within Microsoft's Azure environment. Microsoft commits, contractually, that prompts and inputs, completions and outputs, embeddings, and training data are not available to other customers; are not available to OpenAI or to the other providers of the models Microsoft sells through Foundry; are not used by those providers to improve their models or services; and are not used to train any generative AI foundation model without the customer's permission or instruction. We are the customer, and we never give that permission.

Microsoft hosts these models in its own environment, and they do not interact with the consumer services operated by the model providers. The models are stateless: no prompt or completion is stored in the model, and prompts and completions are not used to train, retrain or improve the base models. Any model fine-tuned for us is available exclusively for our use, is encrypted at rest, and can be deleted by us at any time.

Abuse monitoring at the infrastructure layer

Microsoft operates a system that detects use of the service in ways that violate its terms. Where its automated systems find indicators of abuse, a sample of prompts and generated content may be selected for review — by automated means in the first instance, and where that is not conclusive, by authorized Microsoft employees who reach flagged data only through controlled workstations and just-in-time approval. Data held for this purpose is stored within the Azure geography we designate, is logically separated from other customers' data, and is not used to train any model.

ABUSE-MONITORING RETENTION: STATE THE PERIOD GIVEN IN MICROSOFT'S TERMS AS AT THE DATE OF PUBLICATION, AND RECORD THE DATE CHECKED. MICROSOFT'S PUBLISHED WORDING ON THIS POINT HAS CHANGED; DO NOT PUBLISH A FIGURE FROM MEMORY.

MODIFIED ABUSE MONITORING: STATE WHETHER WE HAVE APPLIED FOR IT AND WHETHER WE HAVE BEEN APPROVED. IF APPROVED, THE STORAGE AND HUMAN-REVIEW PROCESS DESCRIBED ABOVE DOES NOT TAKE PLACE, AND THIS PARAGRAPH IS REWRITTEN TO SAY SO. IF NOT APPROVED, THE PARAGRAPH ABOVE PUBLISHES AS IT STANDS.

Where the processing happens

DEPLOYMENT TYPE — REGIONAL, DATA ZONE OR GLOBAL — AND THE DESIGNATED AZURE GEOGRAPHY: TO BE COMPLETED BY DANIELA PEÑA.

With a regional deployment, prompts and responses are processed in the geography we designate. With a data-zone deployment, they may be processed anywhere within the designated zone. With a global deployment, they may be processed in any geography where the model is deployed. In every case, data stored at rest — including anything held for abuse monitoring — remains in the geography we designate, and Azure's data-processing and compliance commitments continue to apply. Data stored at rest in the service is encrypted with AES-256 by default.

Human oversight

AI tools are used solely for educational, preservation and access purposes. Their output is probabilistic and can be incomplete or wrong. All AI-generated output is subject to human curatorial oversight and review, and output that would alter a collection's description of record passes through human review before it does so. We do not use AI to make automated decisions that produce legal or similarly significant effects on individuals. An interactive avatar is a machine-generated representation and is not the person it represents.

6. Our legal grounds, purpose by purpose (United Kingdom and European Union)

Where UK or EU data protection law applies, the basis on which we process personal information depends on what we are doing with it.

Purpose Information used Legal basis How long
Responding to you and maintaining our correspondence Name, email address, institution, and the content of your correspondence Our legitimate interest in replying to a person who has written to us RETENTION PERIOD
Operating, securing and maintaining the Site IP address, device and browser information, server logs Our legitimate interest in the security and integrity of the Site LOG RETENTION PERIOD
Providing a discovery tool or avatar you choose to use The content of your query or conversation, the collection it concerns, and the technical record of the request CONSENT OR LEGITIMATE INTEREST — DEPENDS ON WHAT IS CAPTURED AND WHETHER THE TOOL IS OPTIONAL PERIOD
Detecting and preventing abuse of the AI services A sample of prompts and generated content, held at the infrastructure layer Our legitimate interest in preventing misuse of the service, and our provider's compliance with its own obligations AS STATED IN SECTION 5
Sending you information you asked to receive Name and email address Your consent, which you may withdraw at any time Until you withdraw consent, then PERIOD
Performing an agreement with you As set out in that agreement Performance of a contract to which you are a party As set out in that agreement
Understanding how the Site is used, in order to improve it TO BE COMPLETED WHEN THE ANALYTICS PROVIDER IS CONFIRMED CONSENT OR LEGITIMATE INTEREST — DEPENDS ON WHAT IS DEPLOYED PERIOD
Meeting our legal and record-keeping obligations As applicable to the obligation Compliance with a legal obligation The period the obligation requires

Cells in capitals are gold fields. They are completed from the Annex before this Policy is published.

7. Cookies and analytics

Cookies that are strictly necessary for the Site to function are set when you visit.

ANALYTICS PROVIDER, EMBEDDED VIDEO AND GALLERY PROVIDERS and any other non-essential cookie or similar technology are not set until you have given consent through an affirmative action, and you may withdraw that consent at any time through CONSENT CONTROL.

You may also refuse or delete cookies through your browser settings; parts of the Site may not work correctly if you do.

8. When we share information

  • with service providers who host, maintain or secure the Site and who process personal information on our behalf, bound by contract to use it only on our instructions;
  • with professional advisers, where necessary and under a duty of confidence;
  • where required by law, court order or a competent authority, or to establish, exercise or defend legal claims;
  • with an institution with whom we are working on a project, where you have asked us to introduce you or have consented to it.

Our processors, in order

Archivista Archives, Inc. operates the platform on our instructions. Microsoft Corporation hosts that platform and runs model inference on it. Each is bound by written agreement to use personal information only on instructions, to keep it confidential, to apply security measures consistent with this Policy, and not to train any AI model on material entrusted to us. Microsoft's processing is governed additionally by the Microsoft Products and Services Data Protection Addendum described in section 2.

A current list of the providers who process personal information on our behalf is available on request from privacy@eternallibrary.ai, and a depositing institution may ask to be told in advance of a change to it.

Legal demands

If we receive a legal demand for material or information deposited with us by an institution, family or community, we will, so far as the law permits: redirect the request to the depositor; notify the depositor before disclosing anything; disclose only what we are legally compelled to disclose; and challenge a demand that is overbroad where we have reasonable grounds to do so.

9. International transfers

We are established in the United States and our service providers may be located in other countries. Where such a transfer requires a transfer mechanism, we use the European Commission's Standard Contractual Clauses, or another approved mechanism, for transfers subject to the European Union GDPR; and the UK International Data Transfer Agreement, or the UK Addendum to those clauses, for transfers subject to the United Kingdom GDPR. Which instrument applies to a given transfer depends on the agreement in force with the provider concerned.

At the infrastructure layer, transfers rely in addition on the mechanisms contained in the Microsoft Products and Services Data Protection Addendum, and on the Azure data-residency options we have selected, which are stated in section 5.

10. How long we keep information

RETENTION PERIODS — STATE THEM.

We keep personal information for as long as needed for the purposes described in this Policy, and thereafter for as long as required by law or reasonably necessary to resolve disputes and enforce our agreements.

DISCOVERY-TOOL AND AVATAR SESSIONS: STATE WHAT IS KEPT AND FOR HOW LONG, INCLUDING ANY CONVERSATION HISTORY HELD BY A STATEFUL FEATURE OF THE INFERENCE SERVICE.

Prompts and generated content held at the infrastructure layer for abuse monitoring are retained as stated in section 5.

A depositor may export its deposited materials, descriptive metadata and attestations in standard, portable formats, on the terms of the relevant deposit agreement. Deletion of a sealed or attested record is itself logged, so that the audit trail stays intact.

11. Your rights

Depending on where you live, you may have the right to ask for a copy of the personal information we hold about you; to have it corrected; to have it deleted; to restrict or object to its processing; to receive it in a portable form; and to withdraw consent you have given.

Residents of the United Kingdom and the European Economic Area may complain to their supervisory authority.

Residents of California have rights of access, deletion, correction and portability, and the right not to be discriminated against for exercising them; we do not sell or share personal information as those terms are defined under California law.

Where the information that concerns you forms part of a collection deposited by an institution, family or community, that depositor decides what happens to the collection under its deposit agreement. Write to us and we will route your request to the depositor and support them in answering it.

Write to privacy@eternallibrary.ai to exercise any of these rights. We may need to verify your identity before we act.

12. Cryptographic records on a public ledger

Where a cryptographic hash or attestation is recorded on a public distributed ledger, only the hash or attestation is recorded. We do not write the underlying archival material to a public ledger, and we do not write information that directly identifies a person.

This prohibition binds every layer. Neither Archivista nor any infrastructure provider writes deposited material, or information that directly identifies a person, to a public ledger on our behalf.

A record written to a public ledger may remain publicly available for as long as that ledger operates, and we may be unable to amend or remove it. Requests concerning personal information held outside the ledger remain subject to this Policy and applicable law.

WHAT EACH HASH IS COMPUTED FROM: STATE IT. A HASH DERIVED FROM MATERIAL THAT IDENTIFIES A PERSON MAY ITSELF REMAIN PERSONAL DATA UNDER APPLICABLE LAW.

13. Security

We apply administrative, technical and physical safeguards appropriate to the information we hold. At our layer and at the platform layer these include:

  • encryption of data in transit (TLS 1.2 or higher) and at rest (AES-256);
  • role-based access control and least-privilege administration;
  • single sign-on and multi-factor authentication;
  • logical isolation of our environment from those of other organizations;
  • authorization controls over changes to a collection's description of record, so that no unsupervised process alters it;
  • cryptographic sealing and tamper evidence for attested records;
  • immutable audit logging; and
  • access by personnel restricted to what is strictly necessary for support and operations, logged and reviewable.

At the infrastructure layer we inherit Azure's physical and platform security controls and its compliance certifications, and data stored at rest in the inference service is encrypted with AES-256 by default.

CERTIFICATION STATUS: PUBLISH ONLY WHAT COUNSEL HAS VERIFIED. AZURE'S CERTIFICATIONS COVER THE INFRASTRUCTURE LAYER AND DO NOT STAND IN FOR OUR OWN OR ARCHIVISTA'S APPLICATION-LAYER ATTESTATIONS.

No method of transmission or storage is entirely secure, and we cannot guarantee absolute security. If a breach affects deposited material or your personal information, we will notify affected depositors without undue delay, and will notify individuals and regulators where the law requires it.

14. Changes to this Policy

We may amend this Policy. The version in force is the one published on the Site, and the date of the most recent revision appears at the head of this page.

We will not reduce your protections quietly or retroactively. Where a change is material we will say so on the Site and, where we hold their contact details, notify depositing institutions by email, and the change takes effect no sooner than thirty days after that notice. No amendment will retroactively enlarge the uses we may make of personal information or of deposited material already held.

If Archivista or Microsoft materially changes a commitment adopted in section 2 or section 5 in a way that affects the protections described here, we will say so on the Site and describe how we are responding.

The version history of this Policy is available on request.

15. Contact

Legacy Studios, Inc., doing business as Eternal Library Foundation
PRINCIPAL OFFICE ADDRESS
privacy@eternallibrary.ai