BioByte

Privacy Policy

Effective

The short version

BioByte holds your medical records so it can show them back to you. That is the whole job, and this policy is mostly about the limits we keep around it.

We do not sell your data. We do not run advertising. We do not run product analytics on it. Nothing enters your record until you have read it and approved it.

You can delete your account from inside the app, and that deletes your documents, your records and your chat history for good.

Who we are

BioByte is an AI powered medical journal. This policy covers the BioByte iOS app and the BioByte website, and it explains what we collect, where it goes, and what you can do about it. If anything here is unclear, write to us at support@mybiobyte.com and we will answer in plain language.

BioByte is not a healthcare provider and not a covered entity under HIPAA. Your records are protected here by this policy and by the law that applies to consumer health data, not by HIPAA. Washington residents have a separate set of rights, set out in our Consumer Health Data Policy.

What we collect

Your account.If you sign up with an email address, we hold that address and a password you never send us in readable form. If you use Sign in with Apple, Apple gives us an identifier for you and whichever email address you chose to share, which may be one of Apple's private relay addresses. We never see your Apple ID password.

Your profile. During onboarding we ask for your name, date of birth, height, weight, blood type, allergies, and, optionally, the sex you were assigned at birth. Every one of those can be left blank, and you can change them later from your profile. The body figure the Journal Body tab draws is a display preference derived from that last field when you have not chosen a figure directly.

Documents you upload. From the iPhone app or the website: the file itself, plus its original filename, its file type, its size, and a checksum we use to notice when you have uploaded the same document twice.

The record we extract from them. Values, units, dates, reference ranges and how confident the extraction was, along with conditions, medications and the timeline entries they hang from. When you correct a value during review, we keep what BioByte originally read alongside your correction, so the record can always show you both.

Apple Health, if you connect it.In the iPhone app you can let BioByte read sleep, steps, workouts, heart rate, resting heart rate, heart rate variability and body weight from Apple Health. During onboarding it can also read your date of birth and height to fill in your profile. BioByte only reads; it never writes to Apple Health. You can turn access off at any time in your iPhone's Health settings.

Photos you track. Photos of a spot or area of skin you choose to follow over time, taken in the app or attached in chat.

Your chat. The questions you ask, anything you attach, and the answers you get back, along with the records each answer cited.

What BioByte learns with you. Facts about you that you state in chat or confirm, and the summaries, suggestions and things to keep an eye on that BioByte writes from your record. A fact BioByte infers, rather than one you said in your own words, waits for your approval the same way an extraction does.

Your research choice. Research search is on unless you turn it off in Vault. We keep whether you turned it off, so we respect that on every question. You can change it in Vault at any time.

A search index over your documents. Document text is split into passages and converted into numeric embeddings so chat can find the right record when you ask about it. The passages and the embeddings live alongside your documents and are deleted with them.

Bug reports, from test builds only. The TestFlight and test builds of the iPhone app have a Report a bug form; the App Store version does not. A report carries what you write, a screenshot of the screen you were on (you can remove it before sending), up to four more pictures you choose, and basic device details: the app version and build, the device model, the iOS version, the screen you were on, and a reference to any matching error report. We store the report with your account identifier and send it to our issue tracker, described below.

Service records. For every AI call we log which model ran, how many tokens it used, what it cost and how long it took. Those rows hold no medical content and no prompt text. We keep an audit trail of what kind of change was made to your record and when, and of when a copy of it left (a summary export or a document download); it holds no record values. We also keep short-lived counters that limit how often something can be done, to stop abuse. Some are keyed by your account identifier, some by a one-way hash of an email address, and some by the IP address a request came from. They hold no name, filename or medical value, and each one is deleted when its time window ends: one keyed by an IP address lasts at most 15 minutes, and none lasts longer than a day.

What we do not collect

We run no product analytics and no advertising. There are no third party trackers, tag managers or advertising pixels on the BioByte site or in the app, and the site enforces a content security policy that permits the browser to talk only to us and to the services listed below. We do not collect your location, your contacts or any advertising identifier. We do not buy data about you, and we do not build a profile of you for anyone else.

Where your data lives

Your account, your profile, your records and your chat are stored in a Postgres database run by Supabase. Uploaded documents are stored in a private Supabase storage bucket. The bucket is not public: a document can only be read through a signed link that our API mints for you and that stops working after five minutes.

The website and the API run on Vercel. Traffic to BioByte is HTTPS only, enforced with HTTP Strict Transport Security. On the web, your session lives in a cookie that JavaScript cannot read; on iOS it lives in the system Keychain.

Who else processes your data

We keep this list short on purpose, and it is complete as of the effective date above.

  • Supabase stores your account, your records and your documents, and sends the sign-in and verification emails.
  • Vercel hosts the website and runs the API, so every request passes through it. It also runs the AI Gateway that every model request goes through.
  • AI model providers, reached only through Vercel's AI Gateway, read the documents you upload, answer your questions in chat, turn document passages into embeddings for search, and write the summaries and suggestions BioByte shows you. The next section says what they receive and what they may keep.
  • Apple handles Sign in with Apple, if that is how you sign in.
  • Sentry receives error and crash reports when something breaks. See below for what those contain.
  • Linear is our issue tracker. It receives the bug reports you send from a test build of the app: what you wrote, the screenshots and pictures attached, and the device details. We do not send your account identifier or email address with them, but a screenshot or what you wrote can still identify you.
  • EMBL-EBI (Europe PMC), NCBI/NLM (PubMed), NLM (MedlinePlus) and the US FDA (openFDA)are third-party recipients of a search query when you ask chat a general research question. BioByte checks the query for the name and date of birth on your profile, for names of people and places written in your recent records, for nicknames and initials, for full dates and birthdays, for addresses and postcodes, and for contact, insurance and ID numbers, including when they are written with common lookalike letters, spelled-out digits or an encoding, and refuses to send it if it finds one. That check cannot catch everything, so avoid typing other people's names or identifying details in a research question; the query can include general clinical details such as a lab value, a medication or an age range. Research search is on unless you turn it off in Vault, and while it is off nothing is sent to them.

We do not sell your personal information or your health data to anyone, and we do not share it for anyone else to advertise to you. If we ever need to add a processor to this list, we will update this page and change the effective date before it starts.

How the AI providers handle your data

Reading a lab report means sending it to a model. When BioByte reads a document you upload, the file is sent to an AI model provider. When you ask a question in chat, your question, the conversation so far, anything you attached, and the parts of your record needed to answer it are sent to one. When a document is indexed for search, its passages are sent to one. There is no way to offer the product without this, so we say it plainly rather than burying it.

Every model request goes through Vercel's AI Gateway and asks for zero data retention. The Gateway sends a request only to a provider that Vercel holds a zero-retention agreement with, and if none of them can serve it, the request fails rather than going anywhere else. Which of those providers serves a given request is the Gateway's choice.

Zero data retention means the provider does not store your request or its answer once it has responded. It has limits, and you should know them: a provider can still hold a request where the law requires it or to check for abuse, and a provider may briefly cache part of a conversation, for a few minutes, to answer the next message faster. We do not use your data to train models, and we do not give it to anyone for that purpose.

What we send is scoped to the task. A question about your records goes with a short background summary of you, so the answer fits: your name, your age, your sex assigned at birth if you gave it, your body figure, some of your current medications and conditions, the spots you track, a few facts you have shared, recent sleep, and anything waiting for your review. It is not your whole history.

Unless you turn research search off in Vault, a general medical question can also search published literature through Europe PMC (run by EMBL-EBI) and PubMed (run by NCBI/NLM), and public health pages through MedlinePlus (run by the National Library of Medicine) and drug labels through openFDA (run by the US Food and Drug Administration), four third-party recipients rather than processors under this policy. BioByte checks the query for the name and date of birth on your profile, for names of people and places written in your recent records, for nicknames and initials, for full dates and birthdays, for addresses and postcodes, and for contact, insurance and ID numbers, including when they are written with common lookalike letters, spelled-out digits or an encoding, and refuses to send it if it finds one. That check cannot catch everything, so avoid typing other people's names or identifying details in a research question; the query can include general clinical details such as a lab value, a medication or an age range.

What error reports contain

When something goes wrong, we send an error report to Sentry so we can fix it. Those reports are filtered before they leave: the error message is removed, request bodies are never included, and query strings are stripped from URLs because a search term can itself be a medical fact. Reports from the iPhone app carry no account identifier at all; reports from our server identify you only by your account identifier. Neither carries your name, email or IP address. What survives is the type of error, where in the code it happened, and which feature you were using.

Your choices

Approve before anything is filed. Extraction output is held as pending. It becomes part of your record only when you say so, line by line, and anything you do not keep is discarded rather than filed.

Correct what BioByte read wrong. During review you can edit a value or its unit before you confirm it. Your correction is what gets filed, marked as yours, with the original reading kept beside it.

Delete a thread. Deleting a thread in the app deletes every message in it, immediately.

Delete your account. In the app, open Vault and choose Delete my vault. This removes your uploaded files and photos from storage and then deletes your account, which cascades through every record we hold for you: profile, timeline, observations, conditions, medications, tracked photos, facts, chat, Apple Health samples, the search index, the bug reports stored with your account and their screenshots, and the rate-limit counters keyed to your account. It is immediate, it ends your sessions, and it cannot be undone. There is no grace period and no way for us to restore it afterwards.

One thing is kept after you delete your account: bug reports already sent to our issue tracker. They carry no account identifier or email address, but what you wrote and the screenshots attached may still identify you. If you want them deleted, write to us and we will delete them.

Export a summary.In the app, Vault's Export everything builds a summary of your record for a doctor: your profile, conditions, medications and lab results, which you can save or share from your phone. It does not include your documents, photos or chat.

Ask us for a full copy. Write to support@mybiobyte.com and we will send you a copy of everything we hold for you, or correct something you cannot correct yourself. We will ask you to confirm you control the account before we do either.

How long we keep it

Until you delete it. BioByte has no automatic expiry: your records are meant to still be there when you need them years from now, so nothing is aged out on a timer, and there is no scheduled purge.

Two details are worth stating because they are easy to assume the other way. If you reject an extraction, we delete what was extracted but keep the original file, so the document can be processed again. If you remove a document from your list, we delete the stored file at the same time rather than later: the entry and the file it came from go together.

Two things follow different rules. Rate-limit counters are deleted when their time window ends, at most 15 minutes for one keyed by an IP address and at most a day for the rest. Bug reports stay in our issue tracker until the BioByte team deletes them, including after you delete your account, without any account identifier attached.

Children

BioByte is not built for children and is not offered to anyone under 18. We do not knowingly collect information from anyone under 18, and if we learn that we have, we will delete the account and everything in it.

Changes to this policy

If we change this policy we will change the effective date at the top of this page. If a change materially affects what happens to data we already hold, we will tell you in the app before it takes effect rather than relying on you to re-read this page.

Contact

Questions, corrections, requests, or a policy you think is wrong: support@mybiobyte.com. See also our Terms of Service and our Consumer Health Data Policy.