Privacy Policy
What Sourceglow processes, where it goes, and how to ask for help or deletion.
Effective date: 2026-11-18Who is responsible
Sourceglow is an independent project operated by Frans Naakka, an individual based in Finland. It is not a registered company. This policy covers the Chrome extension, source-searching service, website, and support communications.
Legal publisher and data controller: Frans Naakka, Finland.
Contact address:
Frans Naakka Heinäsuonraitti 65a A4 37830 Akaa FinlandSupport and privacy email: sourceglow.support@gmail.com. The canonical website is www.getsourceglow.com.
This policy describes the beta service and the choices available to account holders, website visitors, and people mentioned in selected public sources.
Beta audience: Free, first-come-first-served beta access for up to 10 eligible adults residing in the EEA or United States. The public website may be viewed globally; that does not grant service access. The website, extension, onboarding, legal documents and support use English. This beta is intended for testers comfortable reading and using English.
Questions and source documents
Sourceglow finds relevant passages in one webpage or PDF you select. It does not independently assess source reliability, guarantee a claim is true, or search the wider internet for corroboration.
After you accept the first-use search disclosure and select Find sources, the extension sends your question and all text extracted from the selected source to its Railway-hosted backend. This includes passage text and identifiers, headings, passage types, PDF page numbers where present, and extraction-limit information. This transfer is not limited to the excerpt displayed in a result.
OpenAI processes your question and source text to find passages. For longer documents, text chunks from across the extracted source and a representation of your question can be processed by OpenAI’s embeddings service before selected passages are reviewed. The resulting embeddings are derived from your content and are not treated as anonymous.
The original PDF file selected from your computer stays in your browser; its extracted text is transmitted for search. Questions, quoted results, explanations, source mappings, and selected PDF references are also held in extension/browser memory during use.
Opening the side panel reads current-tab information and can check service, session, and beta-account status. It does not itself send webpage text for a search. Signing in and checking beta access can create or update account and membership records before you accept the document-search disclosure.
The search request does not separately include the source URL, title, filename, cookies, or a browsing-history feed. Those details, or other personal information, may nevertheless appear inside the extracted source text or your question.
Browser and source-site processing
The extension reads the current tab’s URL, title, and identifiers to display the selected source, extract its text, and navigate to results. Sourceglow does not operate a background browsing-history collection feature.
For an online PDF, Chrome contacts the source website to obtain the file. These requests may include that website’s cookies, normal connection information, and redirects. The current implementation also permits HTTP source documents; only the connections to Sourceglow’s backend and authentication services are required to use HTTPS.
A PDF highlight may place a short verified source quotation in a URL text fragment. It can be visible in the address bar, copied links, or browser-managed records. Browser history, sync, downloads, and the source website’s handling are separate from Sourceglow’s application database.
The beta is for adults aged 18 or older. Use ordinary, publicly available, non-sensitive informational, research, educational, governmental and technical webpages and PDFs. Incidental ordinary personal information, such as a published author's name, may be present.
Do not submit sensitive personal information, private or confidential documents, or accidentally exposed personal documents. People profiling, personal-data dossiers and questions designed to investigate individuals are prohibited. Sourceglow does not automatically verify age or detect prohibited content. A document being reachable online does not make accidentally exposed personal material permitted.
Public sources can still contain authors’ names or other personal information. Public availability does not remove privacy obligations, and your agreement does not establish a legal basis for processing somebody else’s data.
Google sign-in and account information
Authentication happens in the extension, not on this website. The beta uses Google sign-in through Supabase Auth. The implemented flow requests openid, email, and profile and uses a code exchange with PKCE through Chrome’s authentication window.
Google can share your account identifier, email address and verification status, name, profile picture, and other basic profile metadata with Supabase. Supabase stores the authentication identity and provides Sourceglow with the account/session information needed to sign you in, link identities, and determine beta access.
The flow does not request Gmail messages, Drive documents, contacts, or calendar access. Your Google password is entered with Google. New beta users sign in with Google. If you need help with an existing legacy email account, contact support; do not email your password or assume automatic email recovery is available.
The extension persistently stores its Supabase session and authentication-flow material in Chrome local storage restricted to trusted extension contexts. This includes Supabase access and refresh tokens, identity/profile data, and PKCE material during sign-in. Unused Google API access/refresh credentials supplied by the SDK are removed before local persistence and when older stored sessions are restored. These access restrictions are not encryption. Authenticated backend requests carry a Supabase access token whose signed claims may contain identity metadata.
Account creation is separate from beta admission. After Google sign-in and the eligibility/Terms acknowledgment, the server automatically enrolls eligible accounts on a first-come, first-served basis while fewer than 10 active seats are occupied. Existing admitted accounts keep their membership and lifetime usage. An installation link or completed sign-in does not guarantee a seat.
If the beta is full, an Auth/profile account can still exist without membership or search access. These never-admitted accounts are covered by the retention criteria below; you can request deletion through the privacy email. No waitlist or invitation list is created.
Before Google sign-in, you self-confirm age 18+, EEA or United States residence, ability to use the English beta, permitted-source use and agreement to Terms. This does not verify your actual age or residence. No birth date, identification document, street address, IP geolocation or country-verification service is used for onboarding. Country is not added to the account profile or local acknowledgment record; the acknowledgment stores only its account/backend-scoped version and time.
Beta, usage, and operational records
The free beta allows up to 10 active participants, each with 100 lifetime reserved search attempts. Once a search is accepted and reserved, a failed, cancelled, or no-match search can still count. A request rejected before reservation does not necessarily consume an attempt. The extension’s account panel shows the applicable allowance.
Supabase stores account identifiers, profile creation/access flags, beta admission method and timestamps, active/revoked status, lifetime attempt counts, monthly usage counts, and request state/timestamp/lease/cost records. Monthly records still exist for accounting; the closed beta’s allowance is lifetime, not a monthly reset.
The database also keeps daily aggregate reserved and charged processing-cost totals. Revoking and later restoring the same beta membership preserves its used attempts. Deleting the Auth account cascades the linked membership and its lifetime counter, which is different from revoking access.
Application usage logs contain request identifiers, model and retrieval strategy information, processing counts, token usage, timings, partial-coverage flags, costs, and settlement/error metadata. Request identifiers can be linked to account records, so these logs are not anonymous.
The reviewed application does not intentionally write full questions or document bodies to its persistent database or usage logs. Infrastructure, authentication, AI, and email providers have their own logging and retention. They may process IP addresses, user agents, authentication/security events, and service metadata.
Why information is processed
Questions and extracted text are processed to provide the source-searching feature you request. Identity and session information provide authentication and account access. Beta membership, usage, and cost records enforce admission and allowances and control service spending. Limited operational/security records help diagnose failures and protect the service. Support correspondence is used to respond to your request.
Purposes and GDPR legal bases:
- Necessary account identity, sessions, beta admission and allowances, and requested searches concerning your own data: performance of the beta service or steps you request before entering it, under Article 6(1)(b). This covers necessary processing; adding a use to Terms does not make it necessary.
- Proportionate security, abuse prevention, rate controls, diagnostic/request records, cost protection, and website delivery/security: the legitimate interests in operating a safe, limited service, under Article 6(1)(f), subject to necessity, balancing, and limited retention.
- Support needed for your beta service: Article 6(1)(b); other ordinary inquiries may use the limited legitimate interest in responding under Article 6(1)(f).
- Handling GDPR rights requests and proportionate verification: compliance with the relevant GDPR obligations under Article 6(1)(c). This does not justify keeping every message or log indefinitely.
Other people’s data in public sources: your beta contract does not supply a contract basis for those people. The assessed legitimate interest is helping a reader locate evidence in a selected ordinary public reference source, with incidental ordinary personal information kept in context. It does not cover investigating individuals, profiling, sensitive information or private documents. The narrow balancing assessment supports this limited purpose subject to its necessity, provider and transparency conditions; it does not establish that all public-source processing is lawful.
Article 14 notice: public availability does not remove the duty to inform people whose data is obtained indirectly. Sourceglow has not established a blanket impossibility or disproportionate-effort exemption. Where individual information is required, the operator must provide it within the applicable time limits. An exception may be used only where its conditions are established and documented, with the required alternative safeguards. Publishing this policy alone is not treated as satisfying an individual-notice duty. People mentioned in sources can contact the operator without an account.
Google sign-in authorizes identity sharing. Terms govern the beta service. The separate first-search action acknowledges the disclosed document transfer; it is not blanket GDPR consent or evidence that you accepted Terms. We do not rely on it to authorize unrelated uses or sensitive third-party processing.
Limited Use commitment: Sourceglow follows the Chrome Web Store User Data Policy, including Limited Use, when handling Google API information and other extension data. Uses are limited to the disclosed passage-finding feature, authentication, allowances, security, and support. No advertising, sale of user data, unrelated profiling, or operator model-training use is part of this beta.
See the Finnish Data Protection Ombudsman’s guidance on legal bases and legitimate-interest assessments.
Service providers and sharing
| Recipient | What it processes and why |
|---|---|
| Railway | Hosts the search backend, receiving questions and all extracted source text; processes temporary retrieval data and application/infrastructure logs. |
| OpenAI | Processes questions and document passages for source search and text chunks/embeddings for longer-document retrieval. It receives extracted text, not the original selected PDF binary. |
| Supabase | Handles authentication, session/identity records, database profiles, beta membership, lifetime and monthly usage, request accounting, and security/audit information. |
| Authenticates your Google account when chosen. Google also provides the Gmail mailbox used for support and privacy correspondence; this is separate from the sign-in scopes. | |
| Vercel | Serves the public website and can process connection, delivery, access, and security metadata. The website demo does not send source questions or documents to Vercel for an AI search. |
Source websites and Chrome also handle normal document retrieval and local navigation as described above. GitHub hosts the project’s source code; the search implementation does not send user documents or questions to GitHub.
Providers can act under different roles for service delivery, their own account administration, and security. We do not promise processing only in the EU or assume every provider action is covered by the same processor agreement.
Processing and international transfers: Sourceglow does not offer EU-only data residency. The listed hosting, authentication, AI and email services can process information outside your country, including in the United States. The operator has reviewed the applicable provider and international-transfer arrangements; Railway's data-processing agreement has been signed. Where transfer safeguards are required, the applicable provider arrangements and legally available safeguards govern the transfer. Contact the privacy email for information about the relevant recipients, countries and how to obtain details of applicable safeguards.
Storage and retention
Extension and browser: question/result/document state is kept during use and cleared or replaced through the extension’s lifecycle. Account changes clear panel state and selected PDF references. Session, beta eligibility/Terms acknowledgment, and separate search-disclosure records persist in local Chrome storage. The acknowledgment stores only a version and time, scoped to the signed-in account and backend; it is not a central acceptance ledger. Signing out does not necessarily remove these acknowledgment/disclosure records or browser-managed history/downloads.
Backend document cache: the implementation stores extracted text chunks and embeddings in a user-separated memory cache, with a production-configured expiry capped at 15 minutes after cache insertion. Entries expire automatically and the cache is cleared on shutdown. Request processing can hold text while work completes; timer scheduling, in-flight processing, and provider handling mean this is not a guaranteed physical deletion deadline.
Cloud account and accounting records: the operator manages retention and verified deletion requests manually. Account deletion covers linked application records; aggregate daily costs are separate. Active searches and cached content require separate handling. Provider-managed security records and backups can remain under their own retention criteria; no immediate erasure of every copy is promised.
Beta retention schedule: the operator uses an active weekly administrative review, not an automatic age-based deletion service or a self-service account portal:
- Never-admitted accounts created for this beta: review for deletion after 7 days without admission; protect pending approved cases and historical accounts.
- Beta accounts and lifetime usage: keep while legitimately needed for this beta, then close within 30 days after the beta ends, or handle an earlier verified deletion request.
- Settled request records: 30 days from creation; closed monthly usage: 30 days after month end; settled daily aggregate cost records: 90 days from the budget day.
- Ordinary resolved support correspondence: 90 days after resolution; minimal rights-handling evidence: 90 days after completion unless a specific legal need justifies otherwise.
Active reservations, unresolved cost settlement, and narrowly justified preservation require separate handling. Weekly reviews can occur after a record reaches its eligibility date; they do not promise exact-time erasure. There is no blanket extension of the application schedule to provider logs, AI records, mail-provider copies, or backups.
Provider retention: Sourceglow requests store: false for OpenAI Responses, but this does not switch off all provider retention or prompt caching. Providers can retain security, abuse-monitoring, operational, or backup records under their applicable service settings and legal criteria. Sourceglow does not promise zero retention or an exact provider-erasure deadline. See OpenAI’s data controls.
Supabase logs/backups, Railway logs, Vercel infrastructure records, and Google/Gmail records have separate settings and retention. Their dashboard visibility windows are not a promise that every retained copy has been erased.
Retention is administered through the active weekly operator review described below; it is not an automatic age-based deletion service. Account, allowance and accounting records are retained only for the described beta, security, support and legal purposes. Verified deletion requests receive individual handling. In-flight processing, necessary preservation and provider-managed logs or backups are handled separately; immediate erasure of every copy is not guaranteed.
Website and support communications
The website has no account system, payment processing, waitlist database, advertising trackers, or analytics SDK. Its fonts and brand assets are served with the site. The illustrative demo uses fixed sample text and predetermined results in your browser, without an AI/backend request.
The website does not set application cookies or use local/session storage for the demo. Hosting and security services can still process ordinary request metadata under their own service and security criteria.
Tracking choices: Sourceglow has no advertising or cross-site tracking feature and does not sell personal information or share it for cross-context behavioral advertising. The application does not change its processing in response to browser Do Not Track signals. Deliberately searching a selected source is different from background browsing-history tracking; providers’ separate network/security processing is described above.
Website hosting: Vercel serves this static website and can process ordinary connection, delivery and security metadata. The website has no application account database or AI-search endpoint; its demo runs locally with sample results. Provider-managed network/security records follow the provider's applicable service, retention and legal criteria, separately from Sourceglow's database and correspondence schedules. No zero-log or exact provider-erasure deadline is promised.
If you email sourceglow.support@gmail.com, the operator and Google’s Gmail service process your sender address, message, attachments, and mail metadata to handle the request. Correspondence is subject to the manual support-case schedule above. Sent messages, Trash, local mail copies, and provider backups can have separate lifecycles.
Please do not send passwords, session tokens, API keys, private source files, or unnecessary sensitive information. For document-specific troubleshooting, start with a public URL or a redacted description. Private documents remain outside the beta’s scope; support does not require you to submit them.
Data-use commitments and security
Sourceglow uses data for its disclosed source-searching feature and essential account, security, allowance, and support operations. It does not sell user data or use it for personalized advertising, unrelated profiling, creditworthiness decisions, or its own model training.
Sourceglow’s use and transfer of user data follows the Chrome Web Store User Data Policy’s Limited Use restrictions. Information obtained through Google APIs is also handled in accordance with the Google API Services User Data Policy. These are Sourceglow’s commitments, not a certification issued by Google.
Human access is limited to what is needed for service security, legal obligations, or support you specifically authorize; private source text is not requested merely for convenience. Transfers to providers must remain necessary for the disclosed feature or another permitted purpose.
The implementation uses HTTPS for backend/authentication requests, server-side credentials, verified account tokens, database access controls, admission/rate/cost limits, and user-separated retrieval caches. These safeguards do not guarantee perfect security. Do not share account credentials; report suspected exposure to the support email.
Your choices, rights, and deletion requests
You choose which source to search and when to start. You can decline the search disclosure, stop using the feature, sign out, or uninstall the extension. Signing out or uninstalling does not delete your cloud account, provider logs, or copies retained by the browser.
To request access, correction, export, account deletion, or other privacy help, email sourceglow.support@gmail.com, preferably from the account’s email address. Describe the request without sending credentials. We may need a proportionate check that you control the relevant identity; a claimed email address alone is not verification.
This contact also accepts requests not to sell your information; Sourceglow does not sell personal data. Applicable US state rights can depend on the law and the processing involved. We handle requests under the applicable requirements and do not limit mandatory rights because this beta is free or small.
There is no self-service cloud-account deletion button. The current deletion process requires operator action. It must address active searches and cached content as well as Auth, profile, beta membership/lifetime usage, monthly usage, and request records. Aggregate costs, security records, and provider backups are handled separately; immediate erasure of every provider copy is not promised.
Removing Sourceglow’s access in your Google Account is separate from deleting your Sourceglow account. It does not erase data already held in the service. Signing out also does not necessarily invalidate an already issued access token before expiry.
Under GDPR, you can request access, correction, erasure or restriction, subject to the applicable conditions. Portability applies where its conditions are met; you can object to processing based on legitimate interests. If a use actually relies on consent, you can withdraw that consent without affecting prior lawful processing. We respond without undue delay, normally within one month; any permitted extension must be explained within that initial month.
You may complain to Finland’s Office of the Data Protection Ombudsman or another competent supervisory authority. See Support & account requests.
Email sourceglow.support@gmail.com for access, correction, export, deletion, restriction, objection or other privacy help, including if you have no Sourceglow account. The operator verifies identity proportionately and handles requests manually; passwords and working tokens are not requested. Cloud deletion, active work/caches and provider retention are separate checks. Applicable legal deadlines apply; there is no self-service cloud deletion button or guaranteed ordinary support response time.
People mentioned in sources and prohibited content
You do not need a Sourceglow account to ask about personal information about you in a source. Email sourceglow.support@gmail.com with a minimal description and, if safe, a public source reference. Do not send a sensitive document, credentials or a full dossier to prove your request. The operator may need proportionate information to locate the processing and verify a request; there is no permanent searchable source archive.
The relevant information can include a published author’s name or other incidental ordinary personal information in a user-selected webpage or PDF and question. It comes from the selected public source or the tester’s question, rather than directly from the person mentioned. The purpose, limited legitimate interest, provider recipients, retention limits, transfers and rights are described in the sections above. An English-comfortable tester’s acknowledgment does not establish that a person mentioned in a source has been informed.
If prohibited content is discovered, the operator will stop further affected processing where reasonably possible, assess necessary containment, deletion and applicable notification obligations, and handle relevant privacy or security requests. Repeated violations may lead to revocation of beta access, subject to applicable notice and statutory obligations. Instantaneous deletion from all providers, logs or backups is not promised. If you discover prohibited content, stop searching it and report the issue privately through Support.
Policy changes and contact
The effective date appears at the top of this page. Material changes to data handling must be reflected in this policy and disclosed in the product before the changed processing begins. Where required, renewed affirmative agreement must be obtained.
Questions about privacy, account access, deletion, or service security can be sent to sourceglow.support@gmail.com.