The 23andMe bankruptcy exposed a question every personal AI user should ask: what happens to my history when the company changes? Secret Network’s role in DEPH’s planned wellness platform puts control over access at the centre of the answer.

When 23andMe entered bankruptcy in 2025, customers faced a question they had not answered by taking a saliva test: who would be responsible for their genetic information after the business was sold?

A court approved a $305 million sale to TTAM Research Institute, a nonprofit founded by 23andMe co-founder Anne Wojcicki. The deal kept data in place under existing privacy protections and preserved customers’ deletion rights, with additional safeguards. It was not an unrestricted handover of DNA profiles to a new buyer. But the future stewardship of deeply personal information had nevertheless become part of a bankruptcy proceeding.

Now consider an AI wellness coach that learns about you for twenty years. Its history reaches beyond a single test into your changing routines, sleep, nutrition and conversations about what you want from life. The more useful that relationship becomes, the more consequential it is to lose control of it.

That is where Secret Network enters the story. Through its confidential-computing infrastructure and DEPH’s planned use of it, Secret is addressing a practical problem: how to let an application work with sensitive records without giving it unrestricted possession of them. An AI that knows you for years only works if you can trust the relationship for years.

Who controls my history when the company changes?

A privacy promise and control over access are different things. If an app keeps a readable copy of your history, deciding who gets to use it depends heavily on that company’s policies, contracts and behaviour. Broad permissions can leave room for reuse long after the original conversation or test. Closing an account does not, by itself, establish what happens to copies already shared.

This is not only a theoretical concern. In 2023, the US Federal Trade Commission finalised an order resolving allegations that BetterHelp shared email addresses, IP addresses and health questionnaire information for advertising after promising privacy. BetterHelp settled without admitting wrongdoing and says it did not share clinical information from therapy sessions. The case illustrates how sensitive information can enter advertising systems even when users believe it is being kept private.

Secret’s relevance is the alternative to relying on a promise alone. In DEPH’s design, the individual’s encrypted record is separate from the applications requesting to use it. Holding the stored data is not enough to unlock it. Access also requires permission for the particular use and verification that the approved software is doing the work.

The distinction matters when a service changes hands: possession of encrypted files is not the same as permission to read them. That does not make company ownership or legal safeguards irrelevant. It makes access control part of the technology, rather than leaving it entirely to the next operator’s choices.

How can a coach know me without taking my whole file?

Imagine a shift worker whose sleep patterns change after moving to daytime hours. A useful wellness coach remembers the old routine, understands the new one and helps her revisit goals that no longer fit. Genetics, epigenetic retests and everyday context give DEPH’s personal-coaching approach its continuity. The purpose is wellness guidance, not medical diagnosis.

That continuity needs a history, but it does not require every service to receive a permanent copy of everything in it.

Secret Network supplies the protected computing environment in DEPH’s planned architecture. In that design, approved software opens only the permitted records inside that environment, performs the authorised work and returns the allowed result. A research institution receives an aggregate or another authorised output instead of a bulk download of personal files. A permission names who can use which information, for what purpose and for how long.

The blockchain does not display your biology or publish the text of those permissions. It holds a cryptographic checkpoint—a record that helps detect attempts to rewrite the system’s earlier history while keeping the underlying information private. Secret separates what needs to be checked publicly from what needs to stay confidential.

This is a direction being built, not a claim that every part is already live. On November 2, SecretRebooted highlighted Secret Zone as confidential infrastructure under development for applications; DEPH’s detailed implementation remains a draft design. The proposition is clear: personal AI should gain useful access without automatically gaining lasting custody.

What happens if I change my mind?

This is the question that gives user control its meaning. Someone may want a coach to use her testing history but decline a research project. She may agree to a study and later withdraw. Permission to help her today must not quietly become permission for every future use.

Under DEPH’s Secret-based design, withdrawing permission blocks further access under that permission and stops affected work already running. Temporary working copies are destroyed. The stop is intended to be enforced by the system, not left as a request waiting in a support inbox.

Revocation stops future access. It cannot recall models already trained on your data and released.

A research result already delivered cannot be taken back either. Secret’s access controls govern entry to the records; they do not reach into every later use of a model or result produced with earlier permission. That is why the decision before training matters as much as the ability to withdraw afterwards. Agreeing to personal coaching should be distinct from agreeing to research or model training.

This limit strengthens the case for controlling access before data leaves its protected setting. Once an unrestricted copy has travelled elsewhere, the original service no longer controls every place it can be reused. Keeping the records in place narrows what is handed over; it does not make the consequences of an authorised use disappear.

The lesson of 23andMe is not that personal data should never support research. It is that the company’s future and your choices about your data are not the same thing. A sale, a new partner or a more capable AI should not turn yesterday’s consent into permanent permission.

For Secret, that is a sharper purpose than privacy as a label: make useful access depend on the person’s continuing authority. The next generation of wellness AI will compete on how well it remembers us. Secret’s opportunity is to make sure remembering us does not mean taking control of our history.