WillpowerLab — Privacy Policy Version 3.7. Effective 15 October 2026. This policy explains what WillpowerLab does with your personal data, who else receives it, how long we keep it, and how you get it deleted. Version 3.7 replaces every earlier version. It says more fully what Will is given to answer you in the Lounge (sections 4 and 7). Version 3.6 described what WillpowerLab now does that earlier versions did not: the AI version of your voice, made by ElevenLabs (sections 3 and 5); talking to Will in the Lounge (section 4); WhatsApp messages when your coach's feedback is in (section 4); one-to-one lessons you book with a coach (section 4); signing in with Google or LinkedIn (section 2); sharing a take with a community (section 4d); and what residents of California should know, now that the service is open to them (section 14). It also asks you again for one choice, Help improve WillpowerLab (section 4a), in fewer words: a yes you gave before version 3.6 does not carry over, so it counts as off until you say yes again. Your conversation with Will is never used for training. No recording of your voice, and no clip of one, is ever copied or sent for training. 1. WHO IS RESPONSIBLE FOR YOUR DATA WillpowerLab is operated by an individual based in Poland. The controller of the personal data described here is: Artur Willoński, operating under the name "WillpowerLab" Poland, European Union Contact: contact@willpowerlab.com (a postal address is provided on request to data subjects and to the supervisory authority) The lead supervisory authority under Article 56 GDPR is the President of the Personal Data Protection Office (UODO), ul. Stawki 2, 00-193 Warsaw, Poland. Given the scale of processing we have not appointed a Data Protection Officer, which GDPR Article 37 does not require of us. Privacy requests are handled directly at the address above. 2. WHAT WE COLLECT Things you give us - Your email address and account details. If you sign in with Google or LinkedIn, we receive your email address and the name on that account. - Your slides and presentation materials. - Audio recordings of you presenting. - Edits you make to your Ideal Text, and your responses to feedback. - What you write to Will in the Lounge. - Your phone number, if you give it to us for WhatsApp messages. - If you book a lesson: your name, the time zone you choose, the time you booked and any note you add. - Your answers about other people's clips, if you listen to them. Things produced from your recordings - Transcripts of what you said. - Measurements taken from the sound of your voice: how much your pitch varies, how loud and how varied your delivery is, how fast you speak, how you pause, and how your energy and pitch move across a sentence. - Your Ideal Text and your feedback. - Whether a fragment you re-recorded after an exercise came out better. - The AI version of your voice, and the practice examples made with it (section 3). Things collected automatically - Technical and security logs: IP address, device and browser information, timestamps, and error records. - The version of the app you used, the country you told us you live in, and the language your app was set to, recorded as part of your agreement to this policy. - On the booking page, the time zone your IP address suggests, used only to show you times in your own zone first. We do not store it; we store the time zone you end up choosing. 3. THE SOUND OF YOUR VOICE We take measurements from how your voice sounds and compare them against your own earlier recordings — never against other people. We use them to help choose which of your own sentences to show back to you. We do not produce a score, a rating, a grade or a verdict about you, and no such thing is shown to you, to a coach, or to anyone else. We do not use your voice to identify you. We do not infer, record or guess your sex, gender, age, health, ethnicity, or any other characteristic about you from the sound of your voice. If the measurements cannot be taken reliably from a recording, we record that they could not be taken. We do not fill in a value. If you turn on the choice in section 4a, some of these measurements, taken from a clip a coach has answered a question about, may also help tune the software that spots speaking patterns for everyone: your pace, your pauses, and how long your words and sentence endings last. What the software keeps is settings learned from many people, not your measurements. Without your yes your measurements are never used beyond your own recordings. The AI version of your voice. From the first take you record, we send up to two and a half minutes of its clearest speech to ElevenLabs, a voice AI company, which makes an AI version of your voice for us. We use it to say one of your own sentences back to you the way you could say it, as a practice example. Only you hear these examples, and every one is labelled "AI version of your voice". Each example is checked by transcribing it again, and we measure your next try to see whether the example helped. The AI version of your voice is made for you alone: it is not used to identify you, not used for anyone else, and not used to train any model. You can delete it at any time in Data & consent: we delete it at ElevenLabs and here, with every example made from it, and we stop making examples. You can turn it on again, and we make it from your next take. Other users never hear your voice unless you lend a particular recording to them yourself (section 4b) or share a take with a community (section 4d). 4. WHAT WE USE IT FOR, AND ON WHAT LEGAL BASIS To provide the service: to store your recording, transcribe it, generate your Ideal Text and your feedback, have a WillpowerLab coach review that feedback, make the AI version of your voice and your practice examples, answer you when you talk to Will, and keep all of it available to you. Legal basis: performance of our contract with you (Article 6(1)(b) GDPR). This processing is what the service is; we cannot provide it without doing this. WillpowerLab is hybrid coaching — automated work and a human coach together — so the coach's review sits inside this basis rather than beside it. That basis covers delivering your coaching and nothing else: it does not cover training models, analytics about you, or advertising. We do not do analytics or advertising. Training happens only with the separate choice in section 4a. Talking to Will. Will is the AI assistant in the Lounge. To answer you, Will is given what you write and your recent messages in the Lounge, including the summary after each take and your coach's word. He is also given your current project: its title, what each of your slides says, your Ideal Text slide by slide, how many takes you have recorded, and whether your coach's feedback is in. If your talk is long, he gets the slides you ask about in full and the others by their titles. Will only ever sees your own project. Your conversation is kept in your account so you can come back to it, and you can clear it at any time. Your conversation with Will is private. Nobody at WillpowerLab reads it, except when we must to keep the service safe and working, when the law requires it, or when you ask us to, for example for help with your account. It is never used for training. Legal basis: performance of our contract with you (Article 6(1)(b) GDPR). WhatsApp messages. If you give us your phone number, we send you a WhatsApp message when your coach's feedback is in, with a link to it. The message does not contain your feedback. Legal basis: your consent (Article 6(1)(a) GDPR), which you give by adding your number. You can remove it at any time in Data & consent, and we stop. One-to-one lessons. If you book a lesson with a coach, we use your name, your email address, the time and time zone you chose and any note you add to arrange it, to send you and the coach a confirmation with a calendar entry and the video-call link, and to take payment for it. Legal basis: performance of our contract with you (Article 6(1)(b) GDPR), and legal obligation (Article 6(1)(c) GDPR) for the accounting records of a paid lesson. To keep the service secure and working: fraud and abuse prevention, debugging, protecting the service and its users, and answering you when you write to us. To do this, the person who runs WillpowerLab can see an account's recordings, transcripts and measurements, and can export the measurements of an account's takes, without the account's name or email address. Legal basis: our legitimate interests (Article 6(1)(f) GDPR). To check that our automated listening is right: from time to time a WillpowerLab coach — a person — hears a short clip of a recording and answers one question about one speaking pattern, for example whether the words sound rushed, without being told what our software found. The answer is used only to check and correct the software that chooses your feedback. The coach does not see your name, your words or anything about you, and the answer is never shown to you or used as feedback. If you have turned on the choice in section 4a, the answer may also be used to tune that software, as section 4a says. Legal basis: our legitimate interest in keeping the tool accurate (Article 6(1)(f) GDPR). You can object to it at any time by writing to contact@willpowerlab.com: from then on no clip of yours is chosen for a check. Objecting costs you nothing else. Practice, and making it personal: choosing a short exercise that fits your recording, keeping the fragment you re-record, and remembering what you have been working on so that later exercises suit you better. Legal basis: performance of our contract with you (Article 6(1)(b) GDPR). Practice is part of what WillpowerLab is, alongside your feedback, so it is no longer a separate choice. Doing an exercise is always up to you: you can skip any exercise, and skipping costs you nothing. To meet our legal obligations: for example accounting records, and responding to lawful requests. Legal basis: legal obligation (Article 6(1)(c) GDPR). If you take a paid plan or pay for a lesson, your payment is handled by Stripe. We never see or store your card details; we hold the record that a plan is active or a lesson is paid, and the invoices we are required to keep. We never use the sound of anyone's recordings to train models: no recording, and no clip of one, is copied or sent for training. Making the AI version of your voice is not training: it makes a voice for you alone. What may be used for training, and only with your yes, is set out in section 4a. Recordings we import from elsewhere are covered in section 4c. Do you have to provide this data? For the parts the service is made of, yes. Recording your voice, having it transcribed, and having a coach able to review your feedback are what WillpowerLab is; there is no version of it that works without them, so if you are not willing to provide them you cannot use the service. Practice is part of the service too, though doing any one exercise is always up to you. Your phone number, a lesson booking and talking to Will are up to you. Special categories of data. A recording of you speaking freely may happen to reveal something sensitive — a health condition audible in your speech, or something you mention while presenting. We do not look for this and we do not infer it. Where such information is present, we rely on your explicit consent (Article 9(2)(a) GDPR), which you give as its own separate agreement on the acceptance screen. It is not bundled with anything else, and a contract can never stand in for it. You can withdraw it at any time; because we cannot process a recording of you speaking freely without it, withdrawing it means we stop providing the service. Withdrawing does not affect anything we did lawfully beforehand, and you keep every right in section 8 afterwards — you can still get a copy of your data and still have it deleted. We do not make decisions about you by automated means that produce legal effects or similarly significantly affect you. 4a. HELPING TO IMPROVE WILLPOWERLAB — OPTIONAL If you turn on Help improve WillpowerLab, we use your practice to train the models that give feedback to every WillpowerLab speaker. Your practice, as the Terms define it, is the words of the passages you practise, the numbers measured from them, and what your coach writes and answers about them; never the sound of your recordings, and never your conversation with Will. The screen asks "Can we make the product so much better for you?", says "We'd train our feedback on your practice.", and you say yes with "Yes, make it so much better." What we use, and what for: - Your practice text and your coach's words. The words of a passage you practised, as the transcript holds them, and what your coach wrote about it: a praise line, a clearer version, the words of an exercise, your coach's own line on one moment of a take, and your coach's word for a whole take. They train the models that write those five kinds of feedback. - A coach's answers about your clips. With your yes, a coach may hear a short moment of a recording, sometimes with its words and never with your name, to answer one question: which moment of a passage most needs feedback, or whether a reworded version of your words sounds surer. Those answers, and the answers from the check in section 4 on whether a speaking pattern such as rushing is there, teach the software that spots speaking patterns and chooses which moments to give feedback on. - Measurements of your practice. Numbers taken from a clip a coach answered about, such as your pace, your pauses and how long your words and sentence endings last, and whether an exercise or a practice example in the AI version of your voice helped when you tried again. They tune the same software, and the order in which exercises that fit you equally well are offered. - Separate copies. While the switch is on, we keep a separate copy of the words of each moment you were shown as feedback, and of a coach's yes or no about how assured it sounded, so that training does not depend on your project still existing. Text and numbers only. No recording of your voice, and no clip of one, is ever copied or sent for training. The AI version of your voice, the examples made with it, and your coach's videos are not part of it either. What is used from before you turn it on: the words of passages already in your account, and your coach's words about them, can be used too. Separate copies are made only from takes you record while the switch is on. Legal basis: your consent (Article 6(1)(a) GDPR). You give it on its own screen, by turning the switch on. It is never on when you sign up, never pre-ticked, and never a condition of using WillpowerLab. Saying no costs you nothing. You can turn it off at any time. Your coach's words are your coach's too. What a coach writes about your practice is personal data about you and about your coach. Our agreement with each coach allows what they write to be used this way; your yes is what allows it for what they wrote about you. The trained models. A model trained this way writes feedback for every WillpowerLab speaker, not only for you. Before a trained model is used we test it: if it repeats eight words in a row from the text of anyone who has turned this choice off, it is not used, and it is trained again without that text. The software tuned on coaches' answers and on measurements keeps settings learned from many people, not your words or your measurements. How long we keep the copies: until you turn the switch off or delete your account. If you delete a project while the switch is on, the training copies made from that project stay until you turn the switch off. When you turn it off: your training copies are deleted, here and at the provider that trains for us, and you are left out of every training that starts afterwards, including one already waiting to start. A model already trained stays and may still be used, provided it passes the test above; the settings our software already learned stay too. If you said yes before version 3.6: that yes covered less than this one, so it now counts as off. We treat it as if you had turned the switch off, as the paragraph above describes, until you say yes again. The record of your choice: we keep the record of when you turned the switch on and off, including after you delete your account, for six years after you turn it off or delete your account, and then delete it. It holds identifiers and timestamps, not your words. Who trains, and where: OpenAI trains the models that write feedback for us. Your training copies, which are text, are sent to OpenAI in the United States, under the European Commission's standard contractual clauses. They are deleted there as soon as each training finishes, and when you turn the switch off. The software that spots speaking patterns, chooses moments and orders exercises is tuned on our own systems; nothing for it is sent to OpenAI. 4b. LENDING A RECORDING TO OTHER USERS — OPTIONAL, PER RECORDING You can let other WillpowerLab users hear one short clip of your voice, one recording at a time. It is off unless you turn it on for that recording, and you can turn it off again at any time. What others hear and do: a clip of a few seconds, with no name, no words on screen and nothing about you. A listener answers one question about how assured the clip sounds, or simply listens to it as an example. If you practised that moment, the practised version may be played the same way. Their answers help settle, together with a coach's, whether that moment belongs in your Voice Album; no listener learns who you are, and you never learn who listened. Legal basis: your explicit consent (Article 9(2)(a) GDPR), given per recording by the switch itself. Turning it off withdraws it for that recording: the clip leaves every place it could be heard at once. Nothing you did before is affected, and nothing else in your account changes. Your own answers as a listener: when you answer a question about someone else's clip, that answer is about their voice and it is also your own data. It stays tied to your account, you can ask us for a copy of it (section 8), and it is deleted when you delete your account. 4c. RECORDINGS WE IMPORT FOR TRAINING This section is for people whose voices are in recordings we bring in from elsewhere. They are not WillpowerLab users. We may import recordings of people speaking that we hold the rights to use, to build a set of examples on which our coaches judge speaking patterns and moments, and on which our software is checked and trained. An imported recording is transcribed by OpenAI, measured as section 3 describes, and heard by WillpowerLab coaches without a name. It is never shown to WillpowerLab users, and the person speaking has no account with us. What we hold: the recording, its transcript, its measurements, and coaches' answers about it. Like everything else used for training, it is never sent to anyone for training as audio. Legal basis: our legitimate interest in coaching software that is accurate and tested on real speech (Article 6(1)(f) GDPR). We do not look for sensitive information in an imported recording, we do not infer it, and we do not use it. How long we keep it: while we hold the right to use it. It is deleted when that right ends, or when the person speaking asks us. Your rights: if your voice may be in a recording we imported, section 8 applies to you. You can ask us whether we hold a recording of you, ask for a copy, and ask us to delete it, and you can object to its use at any time by writing to contact@willpowerlab.com. An objection takes your recording out of training and checking. 4d. SHARING A TAKE WITH A COMMUNITY — OPTIONAL, PER TAKE Before your next take, you can choose to share it with a community: the general community, which is every WillpowerLab user, or a community you joined with its pass code, which is only its members. It is not shared unless you choose a community for that take, and you can choose "Only me". What others hear and do: the short clips of that take, with no name, no words on screen and no measurements; listeners see only the community's name. A listener may answer a question about how a clip sounds. Those answers are kept with the clip and may help settle, together with a coach's, whether a moment belongs in your Voice Album. They are never used to train models on their own. Legal basis: your explicit consent (Article 9(2)(a) GDPR), given for each take by choosing a community. You can stop sharing a take at any time: its clips leave every community at once. Nothing else in your account changes. 5. WHO ELSE RECEIVES YOUR DATA AI providers OpenAI receives: - the audio of your recording, in order to transcribe it; - parts of your transcript and related text, in order to generate your Ideal Text and your feedback; - what you write to Will, your recent Lounge messages and your project as described in section 4, in order to answer you; - each practice example made in the AI version of your voice, in order to check by transcribing it that it says your words. ElevenLabs receives: - up to two and a half minutes of audio from one of your takes, in order to make the AI version of your voice; - one of your own sentences at a time, in order to say it in that voice as a practice example. OpenAI and ElevenLabs are the AI providers that receive your audio or your transcript: OpenAI to transcribe it, check it and write your feedback and Will's answers, ElevenLabs to make the AI version of your voice. Each transfer is made under a short-lived internal permit that records which operation it was for and the minimum data it was allowed to carry. We do not authorise OpenAI or ElevenLabs to use your content to train their own models. If you turn on the choice in section 4a, your training copies, which are text, are processed by OpenAI to train WillpowerLab's models, and by no one else. OpenAI also transcribes the recordings we import for training (section 4c). Infrastructure providers - Cloudflare (R2): storage of your audio recordings, of your practice examples, and of your coach's video answers. - Supabase: our database, sign-in, and audio storage on a fallback path. - Railway: hosting of the application. Railway is our processor for what it hosts on our behalf, and a controller in its own right for its own account and service-usage records about us as its customer. - Resend: transactional email, including lesson confirmations. - Vercel: hosting of the website. - Sentry: error reports, stored in the European Union. Messaging - Meta (WhatsApp), if you give us your phone number: your number and the message that your coach's feedback is in, with its link. Meta also handles that message as a controller for its own service, under WhatsApp's terms. Signing in - Google or LinkedIn, if you choose to sign in with them. They tell us your email address and name; they decide for themselves, under their own terms, what they do with the fact that you signed in. Payments - Stripe, if you take a paid plan or pay for a lesson. Stripe is not our processor: it decides for itself how it uses payment data, as a controller in its own right, under its own terms. We never see or store your card details. Human coaches A WillpowerLab coach — a person — may listen to your recordings in order to review the feedback you were given and to prepare practice for you. The review is part of the service rather than an extra you switch on, because WillpowerLab is hybrid coaching; the Terms say so in section 1 and section 11, and a written review that comes back to you is paid for in tokens. A coach who reviews your recording does not see the voice measurements described in section 3. Coaches are bound to confidentiality and see only what a review requires. A coach may answer you in writing or with a video; a video answer is for you. If you book a lesson, the coach receives your name, your email address, the time and any note you add. A coach may also hear a short clip of a recording to check our automated listening, as section 4 describes, unless you have objected to it. If you turn on the choice in section 4a, a coach may also hear a moment of yours, and sometimes read its words, to answer a question that teaches our software, as section 4a describes. Other users If you lend a recording (section 4b) or share a take with a community (section 4d), other WillpowerLab users hear those clips, without your name or your words. Nobody else does. Others We share data with professional advisers, and with authorities where the law requires it. We do not sell your personal data and we do not share it for advertising. 6. TRANSFERS OUTSIDE THE EEA Some of the providers above process data in the United States. Every one of those transfers is made under the European Commission's Standard Contractual Clauses or the EU-US Data Privacy Framework, as each provider's data processing agreement sets out. - OpenAI: Standard Contractual Clauses. Its agreement appoints OpenAI Ireland Limited to process data from the EEA and Switzerland. Training copies (section 4a) are transferred to OpenAI in the United States under the same clauses. - ElevenLabs: the EU-US Data Privacy Framework, and Standard Contractual Clauses, which its data processing agreement incorporates. - Cloudflare: Standard Contractual Clauses. Our storage carries an Eastern Europe location hint, which is a preference rather than a guarantee. - Supabase: Standard Contractual Clauses and a UK addendum. Your data is stored in Ireland. - Railway, Resend, Vercel: Standard Contractual Clauses. - Meta (WhatsApp): the EU-US Data Privacy Framework, under WhatsApp's Data Transfer Addendum, which allows Standard Contractual Clauses instead. - Sentry: error reports are stored in the European Union. You can ask us for a copy of the safeguards in place. 7. HOW LONG WE KEEP IT We keep what you give us for as long as your account is open. When you ask us to delete something, or when you delete your account, we delete it — section 9 describes what that involves and what survives it. Recordings, transcripts, Ideal Text, feedback, voice measurements, practice attempts and account records: kept while your account is open, and deleted when you ask or when you close it. Voice measurements are deleted together with the recording they came from and are never kept after it. A coach's answer in a blind check is deleted together with the recording it was about, and never kept after it. Your coach's video answers: kept while your account is open, including after you delete the take they answer. They are deleted when you ask us or when you close your account. The AI version of your voice and the practice examples made with it: kept until you delete it or close your account, then deleted here and at ElevenLabs. Whether an example helped you is kept with your practice, like your other practice attempts. Your conversation with Will: kept while your account is open, and deleted when you clear it or close your account. Your phone number: kept until you remove it or close your account. A lesson you booked: kept while your account is open, and deleted when you close it, except the record of a paid lesson, which we keep with our accounting records for as long as the law requires. A recording you lent stops being heard by others the moment you turn the switch off, and is deleted with the recording like everything else. A take you shared stops being heard by others the moment you stop sharing it, and is deleted with the take like everything else. Your answers about other users' clips are kept while your account is open and deleted when you delete it. Training copies, if you chose to make them (section 4a) — text only, never audio: the words of practised passages, what your coach wrote about them, and a coach's yes or no about a moment. Kept until you turn the choice off or delete your account, then deleted here and at OpenAI. A coach's answer about a moment of yours given for section 4a, and the two versions of your words it compared, is deleted with the recording it was about, and when you turn the choice off. A trained model, and the settings our software learned from many people's answers and measurements, are kept until we replace them. They hold no copy of your text or your measurements, and a model that fails the test in section 4a is not used. Recordings we import for training (section 4c): kept while we hold the right to use them, and deleted when that right ends or when the person speaking asks us. Record of your agreement to this policy: kept for as long as we may need it to show that your recordings were processed lawfully, and then deleted. Record of your training choice: kept, including after you delete your account, for six years after you turn the choice off or delete your account, and then deleted. We are not quoting fixed retention periods here. A period in a privacy policy is a promise that something is deleted on a clock. We would rather tell you what we actually do than name a number we do not yet enforce. When the automatic schedule is in place we will publish it, and you will be asked to accept the version that describes it. What OpenAI holds, separately from us OpenAI keeps the audio and text sent to its API for up to 30 days, to check for abuse, and then deletes it. On our current plan that period cannot be switched off. It is not used to train OpenAI's models. 8. YOUR RIGHTS You have the right to: - get a copy of the personal data we hold about you; - have inaccurate data corrected; - have your data deleted; - restrict how we use your data, or object to our using it on the basis of our legitimate interests; - receive your data in a portable format; - withdraw any consent you have given, at any time, without affecting what we did lawfully before you withdrew it. To exercise any of these, contact contact@willpowerlab.com. We respond within one month. If we need longer because a request is complex, we will tell you within that month and explain why. If you are unhappy with how we handle your data, you can complain to your local supervisory authority. In Poland that is the President of the Personal Data Protection Office (Prezes Urzędu Ochrony Danych Osobowych), ul. Stawki 2, 00-193 Warszawa. 9. DELETING YOUR DATA — WHAT ACTUALLY HAPPENS We want to be precise about this rather than reassuring. When you ask us to delete your data, we build an inventory of everywhere it is: database records, stored audio files, transcripts, generated documents, queued processing jobs, records of what was sent to providers, and coach copies where they exist. We then delete them and record proof of each deletion. If our process reaches something it cannot resolve on its own, it stops rather than guessing, and a person completes it. That is deliberate: it is better for the process to halt than to report a deletion it did not actually make. It means some deletions are finished by hand. We finish them within the one-month period in section 8, and we tell you when it is done. Three things survive deletion: - Records that prove your recordings were processed with your agreement. These contain identifiers, timestamps and fingerprints — not your recordings, your transcripts or anything you said. We keep them because they are the evidence that we handled your data lawfully, and we delete them when we no longer need them for that. - The record of your training choice (section 4a): when you turned it on and off. It holds identifiers and timestamps, not your recordings or anything you said, and we keep it for six years after you turned it off or deleted your account, then delete it. - Anything we must keep by law, for example accounting records, including the record of a paid lesson. Your training copies, and the answers a coach gave about your moments for section 4a, do not survive an account deletion: they are deleted with everything else. Audio already sent to a provider is deleted at the provider as part of the same process, and we record the outcome. The AI version of your voice is deleted at ElevenLabs before anything else. 10. SECURITY Your recordings are stored with access restricted to the systems that need it. We verify each stored recording against a fingerprint of its exact bytes so we can tell that what we hold is what you uploaded and that nothing has been altered. Internal access is limited and logged. Records that prove how your data was handled cannot be edited or deleted by the application. No service is perfectly secure. If a breach affects your rights, we will notify you and the supervisory authority as the law requires. 11. CHILDREN The service is for adults aged 18 and over. We do not knowingly collect data about anyone under 18. If you believe we hold data about a child, contact us and we will delete it. 12. CHANGES TO THIS POLICY When we change this policy we publish a new version with a new version number. The exact words of each version are fingerprinted, so you will be asked to accept the new version before continuing to use the service, and your record shows exactly which version you agreed to and when. 13. CONTACT Artur Willoński, operating under the name "WillpowerLab" Poland, European Union Contact: contact@willpowerlab.com (a postal address is provided on request to data subjects and to the supervisory authority) 14. IF YOU LIVE IN CALIFORNIA This section adds to the rest of this policy for residents of California. Everything above applies to you too, including your rights in section 8. What we collect, and why: the categories of personal information we collect are those in section 2: identifiers, such as your email address, your name and your phone number; audio recordings of your voice, and the AI version of your voice made from them; what you write, such as your Ideal Text and your messages to Will; and internet activity, such as technical and security logs. We collect them from you and from your use of the service, for the purposes in section 4, and keep each for as long as section 7 says. Sensitive personal information: the recordings of your voice, and anything sensitive they happen to reveal. We use it only to provide the service you asked for, and, if you turn on the choice in section 4a, to train models as that section describes. You can turn that choice off at any time. We do not sell your personal information, and we do not share it for cross-context behavioral advertising. We do not knowingly collect personal information from anyone under 18. Your rights: to know what personal information we collect, use and disclose, and to get a copy of it; to have it deleted; to have it corrected; to limit our use of your sensitive personal information; and not to be treated differently for using any of these rights. To use them, write to contact@willpowerlab.com. Someone you authorise may also write for you. We confirm who you are before we act on a request, and we answer within 45 days.