The short version
- While the side panel is open, Jobi checks the tab you are on to see whether it is a job posting. That check runs on your device and the page text is not uploaded.
- Page text is only sent to our server when you ask for something that needs it, such as analysing a job description or writing a cover letter.
- Jobi does not record your browsing history, and it never sends us the address of a site you visit.
- Settings → Privacy has the controls: download everything in your account, reset Jobi's memory, wipe this device, turn page access off, delete your account.
- Your data is not used to train AI models, ours or Anthropic's.
- No advertising, no ad trackers, and your data is never sold. We do keep the limited product analytics described in section 2.
- The database lives in the EU, and Stripe handles payments so card details never touch Jobi.
A friendly summary, not the policy itself. The full policy below is what applies.
1. Who we are
Jobi ("we", "us") is a job-search companion delivered as a browser extension. The data controller is Mahin Sattar Jahan trading as Jobi, United Kingdom. For any privacy question or to exercise your rights, contact support@jobi.life. We are registered with the UK Information Commissioner's Office (ICO), registration reference ZC200184.
2. What data we process
What you give us
- Your name, email and profile picture, from Google sign-in.
- Your professional profile: headline, target role, desired salary, location, bio, experience, education, certifications, skills, and portfolio or LinkedIn links.
- Documents you upload: CVs, cover letters and portfolio materials. We store the text extracted from them.
- Jobs you save, and the cover letters and CVs Jobi generates for you.
- Your chat messages.
What Jobi generates
- A memory of you, written by the AI, that makes Jobi's help personal rather than generic.
- An activity log of the roles you have applied to or viewed.
- Generated content: cover letters, tailored CVs, email drafts, company research, job-description analyses and interview transcripts.
- The scored debrief from each mock interview, which stays in the conversation.
- Your coaching plan, and your progress through it.
- A record of how familiar Jobi's tone with you should be.
What we collect automatically
- Counters of AI actions per day and interviews per month, used for fair-use limits.
- Your billing status.
- The product analytics described further down this section.
- Your language preference, British or American spelling, which is sent with any request that produces prose.
Company logos are fetched from logo.dev using the company name or domain. On your device, the extension keeps a working copy of your data. Section 11 lists it in full.
Reading the page you are on
The automatic half. Whenever the side panel is open, Jobi inspects the tab in front of you to work out whether it is a job posting. It does this when you switch tabs, when a page finishes loading, when the panel regains focus, and once when the panel opens. It reads only structured job details: job title, company, salary, location, the page URL and title, and the site's favicon. It finds them in the page's job-posting metadata and by scanning the first 12,000 characters of visible text for a salary pattern.
All of that stays on your device. None of that text is uploaded. The only thing that leaves your browser is a single analytics event naming the job platform, described under "Analytics" below. This automatic reading happens only if you have granted page access. That permission is off by default, and you can revoke it at any time in Settings → Privacy.
The half you ask for. Some of the things you ask for need the page: analysing a job description, writing a cover letter, tailoring your CV, running a job-aware mock interview, or answering a question about the page you have open. For those, up to 30,000 characters of that page's text are sent to our server and on to Anthropic. If the tab is a PDF, the PDF file itself is uploaded, up to 10 MB. Jobi fetches that PDF as the signed-in you, so it can include a PDF that sits behind a login.
The distinction worth holding on to: looking for a job posting happens on its own and stays on your device. Sending a page to the AI happens because you asked for something.
Jobi never reads any tab other than the one in front of you, and it never reads anything while the side panel is closed.
What travels with a chat message. When you send Jobi a message, the request carries more than the message itself. It also includes:
- The conversation so far, up to a 120,000-character budget
- The full text of anything generated in that conversation: cover letters, tailored CV bullets, email drafts including subject lines, and job-description analyses
- A short profile string: your name, target role, salary expectation and location
- The page context, if a page was read, up to 8,000 characters
- The screen you are on, and today's date
- Your streak and daily-goal figures
- How many jobs you are tracking, and how many sit at each stage
- The company and job title of every job at interview stage
- The company, title and stage of the six jobs you added most recently
Put plainly: the employer names and role titles in your tracker travel with every message you send, so that Jobi can answer questions about your search. All of it goes to Anthropic, which provides the model that writes the reply.
Jobi's memory of you
Separately from answering you, Jobi uploads the whole conversation in order to build a memory of you. This fires about thirty seconds after you stop typing, and again when you close the panel. The result is a distilled written picture of you, stored on your profile and added to later conversations so that Jobi's help is personal rather than generic. That picture is what "Reset Jobi's memory" erases.
Voice and dictation
Speech-to-text uses your browser's built-in Web Speech API. In Chrome that means your audio is sent to Google's speech service to be transcribed. Jobi never receives or stores audio. Your microphone is also read to drive a decorative level meter, and that reading never leaves the page.
What our server receives is text. For dictation: only the transcribed text, plus the last 240 characters of what you were writing for context, and only when you have cleanup switched on. The "exactly what I said" mode never contacts our server at all. For mock interviews: only the text transcript, the job context and the settings you chose. Jobi's own interview questions are turned into speech by ElevenLabs, so the question text goes out, not your voice.
Analytics
Jobi keeps two kinds of record here, and they are not the same.
Product events. When the automatic scan concludes that a page is a job posting, Jobi sends one event so that we know which job platforms to support. It contains the platform name, taken from a fixed list (Greenhouse, Workday, Lever, Indeed and about thirty others), or the word "other" for anything not on that list. It never contains the site's address, the page URL, the job, or any identifier for the employer. Product events in general are limited to an allowlist of event names carrying fixed values, enforced on our server. They contain no company names and no job titles. An earlier build did send us the domain for sites that were not on the list; it no longer does. Jobi does not record your browsing history, and it never sends us the address of any site you visit.
Signals. When you accept a document Jobi has generated, we store a short line of free text against your account. It looks like "Cover letter:" followed by the company name, or "Email:" followed by the subject line. These are stored on your account and feed the AI memory that personalises Jobi. They are what lets Jobi tell you that you applied to three fintechs this week. They are erased by "Reset Jobi's memory" in Settings → Privacy.
Saved application answers
Jobi's autofill feature was removed before the beta, and it is now gone entirely. Jobi does not save application answers, nothing updates them, and nothing reads any that an earlier build saved. Leftover saved answers from an earlier build sit only on your device. They are erased by Clear stored data in Settings → Privacy, or when you sign out. Your CV is different: the file goes to Anthropic, and the extracted text is stored on our server rather than on your device.
Feedback you send us
Settings has a Feedback screen. What you write there, any screenshots you attach, and the address you give for a reply are sent to us and land in the inbox of the person who builds Jobi. The address is optional and is only used to reply to you.
Screenshots are yours to choose. Jobi never takes one by itself, and it cannot see your screen. Please check what is in a picture before you attach it: a screenshot of a job page may carry your name or the employer's.
We keep feedback so we can act on it and see whether a fix worked. Ask us at support@jobi.life to delete something you sent.
Jobi never reads your email and has no access to your mailbox. When you ask Jobi to draft an email, the draft appears in chat as text for you to copy. Sending it is entirely yours, in your own mail app.
Special-category data
Your CV or cover letters may contain special-category data (for example health, ethnicity, religion, trade-union membership). We do not ask for it. Where your uploads contain it, we process it only to provide the service you asked for, on the basis of your explicit consent given when you upload the document. Please avoid including special-category data you don't want processed. Beyond that, Jobi deliberately never fills in or stores gender, ethnicity, religion, disability, veteran status or sexual orientation.
At a glance, here is what Jobi collects, when, and where it goes:
| Data | When | Sent to |
|---|---|---|
| Name, email, avatar | Sign-in (Google) | Supabase (auth + profile) |
| Job title, company, salary, location, page URL and title, favicon from the tab you are on | Automatically while the panel is open, if you granted page access | Nowhere. It stays on your device |
| Job platform name, or the word "other" | Once, when a page is detected as a job posting | Railway API (analytics). Never the URL or the domain |
| Page text (≤30,000 characters), or the PDF file (≤10 MB) | Only when you ask for something that needs the page, after granting access | Railway API → Anthropic |
| Your message, conversation history, generated documents, profile string, and tracker summary including company and job titles | Every chat message you send | Railway API → Anthropic |
| The whole conversation, to build Jobi's memory of you | About 30 seconds after you stop typing, and when the panel closes | Railway API → Anthropic |
| CV, cover letters, signature images | You upload them | Railway API → Anthropic (we store the extracted text) |
| Microphone audio | While you dictate or run a mock interview | Google's speech service, via your browser. Never to Jobi |
| Transcribed text (plus the last 240 characters for context) | Dictation, and only with cleanup switched on | Railway API → Anthropic |
| Interview question text | During a mock interview | ElevenLabs (voice) |
| Saved links, profile fields, job list | You enter them | Supabase |
| Company name or domain | To show a company logo | logo.dev |
| Your feedback, any screenshots you attach, and the address you give for a reply | Only when you send feedback from Settings | Railway API, then Resend, which delivers the email |
| Payment details | If you subscribe | Stripe (Jobi never sees card data) |
3. Why we process it, and our legal basis
| Purpose | Legal basis |
|---|---|
| Provide the core service | Contract (Art. 6(1)(b)) |
| Process your uploaded documents/CVs (incl. any special-category data) | Explicit consent (Art. 6(1)(a) + 9(2)(a)) |
| Read the tab you are on to detect job postings | Consent (Art. 6(1)(a)), given by granting page access and withdrawable at any time |
| Build Jobi's memory of you, coach you, and judge how familiar its tone should be | Legitimate interests (Art. 6(1)(f)) |
| Read and act on feedback you send us | Legitimate interests (Art. 6(1)(f)) |
| Product analytics, to see which job platforms to support | Legitimate interests (Art. 6(1)(f)) |
| Secure the service, prevent abuse, enforce fair-use limits | Legitimate interests (Art. 6(1)(f)) |
| Take payment and keep tax records | Contract + Legal obligation (Art. 6(1)(b)/(c)) |
We do not use your data for advertising, we do not sell it, and Jobi carries no ad trackers and no cross-site tracking. The product analytics we do keep are the ones described in section 2, and they never include the sites you visit.
4. Who we share it with (sub-processors)
We use trusted providers that process data only on our instructions:
- Supabase, accounts and data, hosted in the EU (Ireland)
- Railway, which hosts our API
- Anthropic (Claude), which provides the models behind every AI output, and which reads your uploaded CV file directly
- ElevenLabs, the interviewer voice, which receives question text only
- PDFShift, which renders your documents as PDFs
- Stripe, payments
- Resend, which delivers the email when you send feedback from Settings. It receives only what you wrote, any screenshots you attached, and your reply address
- logo.dev, company logos
- Google, sign-in, and speech-to-text through your browser's built-in speech recognition
Jobi links out to job sites and to the companies you research. Those sites have their own privacy policies, and we are not responsible for what they do.
5. Training AI models
We do not use your data to train AI models. Not your CV, not your messages, not the documents Jobi writes for you. They are used to give you the service you asked for, and for nothing else.
Anthropic provides the models behind Jobi's output. Under the commercial terms we are on, what Jobi sends is not used to train Anthropic's models either. Every provider in section 4 processes your data only on our instructions.
6. International transfers
Our database is in the EU (Ireland). Some providers are in the United States. Where we transfer data outside the UK/EEA, we rely on appropriate safeguards (the UK International Data Transfer Addendum / EU Standard Contractual Clauses). Ask us at support@jobi.life for details.
7. How long we keep it
We keep your data while your account is active. You can delete any item, reset Jobi's memory and activity log, or delete your whole account at any time from within the app. Deletion is immediate and permanent in our systems, and deleting your account also removes your customer record at our payment processor. Accounts inactive for 24 months are deleted. Your chat history and the documents Jobi has written stay until you delete them or reset Jobi's memory. Payment records are kept by Stripe for as long as tax law requires.
8. Your rights
You have the right to access, rectify, erase, restrict or object to processing, the right to data portability, and the right to withdraw consent. Most of that is a button rather than an email. In Settings → Privacy you will find:
- Download my data, which gives you a file containing everything in your account
- Reset Jobi's memory, which erases the AI's learned picture of you and your activity log
- Clear on-device data, which wipes the copy cached on this device
- Job-page access, the permission switch behind the automatic job detection in section 2. It is off until you turn it on, and you can turn it off again whenever you like
- Use the current tab in chat, which controls whether free-form chat may read the page you have open
- Delete account, which is immediate and permanent and erases everything we hold on our servers, including your customer record at Stripe
- Make a privacy request, which emails support@jobi.life for anything else, such as restricting processing or objecting to it
We answer a rights request within a month, and it costs you nothing.
Wherever you live, those controls are yours. If the law where you are gives you more than this policy describes, write to support@jobi.life and we will honour it.
You can also complain to the ICO (ico.org.uk).
9. Automated decisions
Jobi writes drafts, scores your practice interviews and suggests what to do next. All of it is advice for you to take, edit or ignore. Nothing Jobi does decides anything about you, and none of it reaches an employer unless you send it yourself.
10. Payments
Payments are handled by Stripe on Stripe-hosted pages. Your card details go straight to Stripe and never touch our servers. We store only a customer reference, your plan, and subscription status.
11. Cookies & local storage
We use no advertising or analytics cookies. The extension keeps a working copy of your data in your browser's extension storage, unencrypted. That copy is:
- Your sign-in tokens, access and refresh, and the profile from Google sign-in, including your email address
- Your current chat history, including any cover letters, email drafts and analyses generated in it
- A cached copy of your tracked jobs: company, title, location, salary, stage and dates
- The profile built from your CV: name, target role, salary expectation, location, bio, education, experience, certifications and skills
- Your saved application answers, which can include your phone number, address, date of birth, salary expectation, notice period and work authorisation
- Your preferences: theme, language variant, voice and dictation settings, and your daily goal
This copy is per device and per install. It is never synced between your devices. It is removed when you use Clear on-device data, when you sign out, or when you uninstall the extension. The text extracted from your uploaded CV is not kept on your device. It lives on our server.
12. Children
Jobi is for adults. You must be 18 or over to use it.
13. Security
We protect your data with encryption in transit and at rest, row-level security so each user can only reach their own data, per-user rate limits, and least-privilege access.
If your data is ever caught up in a breach, the law gives us 72 hours to report it to the ICO, and we will tell you as well if it puts you at real risk.
14. Changes
We may update this policy. When we do, we change the version and date above, and for a material change we ask you to re-acknowledge it in the app before you carry on using Jobi. We will not apply a material change to data we have already collected without asking you first.