The architecture is the security model

Every tool here runs inside your browser tab. FFmpeg compiled to WebAssembly handles video and audio, pdf.js and pdf-lib handle PDF, JSZip handles PPTX, and the Canvas API handles images. The site itself is a static bundle of HTML, CSS, and JavaScript — there is no application server, no upload endpoint, no file storage, and no queue.

That removes whole categories of risk rather than mitigating them. A file that is never transmitted cannot be intercepted in transit. A file that is never stored cannot be retained past its usefulness, exposed by a misconfigured bucket, or produced under legal compulsion. There is no "we delete your files after an hour" promise here because there is no hour and no deletion — the file only ever existed in memory on your own machine.

You can verify this rather than trust it: open developer tools, select the Network panel, and process a file. The engine libraries load; your file goes nowhere.

No user database to breach

There are no accounts, no passwords, no email list, and no stored personal data. The most damaging thing an attacker could extract from this site's storage is HTML that is already public. Credential stuffing, password resets, session hijacking, and account takeover are not risks we have hardened against — they are risks that do not exist, because there is nothing to log into.

Nothing to install

There is no desktop application, no installer, and no browser extension. You never grant this site filesystem access, and it never runs code outside the sandbox your browser already enforces on every page you visit. Anything claiming to be an official NotebookLM Remover download or extension is not ours.

Files reach the tools only through a file picker or a drag-and-drop, both of which hand the browser a single file you selected deliberately. The page cannot reach anything you did not choose.

Transport and headers

Every page is served over HTTPS, with HSTS set so browsers refuse to fall back to plain HTTP. The site is hosted on Cloudflare Pages behind its CDN and DDoS protection. A few response headers are worth naming, because they are checkable:

Content-Security-Policy — restricts script, style, font, and connection sources to an explicit allowlist, so injected third-party code has nowhere to load from.
X-Frame-Options: DENY — the site cannot be framed, which rules out clickjacking a file into a hidden UI.
X-Content-Type-Options: nosniff — no MIME-type guessing.
Referrer-Policy: strict-origin-when-cross-origin — outbound requests do not leak full URLs.
Permissions-Policy — camera, microphone, geolocation, and payment are switched off for this origin entirely.
Cross-Origin-Embedder-Policy and Cross-Origin-Opener-Policy — required to enable the shared memory FFmpeg needs, and they also isolate this tab from cross-origin pages.

Third-party code

The processing libraries — the FFmpeg WebAssembly core, pdf.js, pdf-lib, and JSZip — are open-source projects, self-hosted on cdn.notebooklmremover.org rather than pulled from a public package CDN. Self-hosting means the code delivered to you is a version we control and can pin, instead of whatever a shared CDN happens to serve that day.

These libraries execute in your browser and receive your file, which is the point: they are what does the work. They are also subject to the same content security policy as everything else on the page, so they have no route to send data anywhere the policy does not allow.

The site additionally loads Google AdSense, Google Analytics, and Google Fonts. Ads render in their own frames and are isolated from the page's data by the browser's origin model — an ad script cannot read the file you dropped in. What each third party does receive is set out in the privacy policy.

Limits worth knowing

Local processing has trade-offs, and pretending otherwise would be dishonest. Your file is handled by your own browser, so the security of the operation also depends on your machine: a compromised device, a malicious extension with permission to read page content, or a shared computer where the downloaded output stays in the Downloads folder are all outside anything this site can control. On a shared or public machine, clear the downloaded file when you are done.

Reporting a vulnerability

If you find a security issue, we want to hear about it. Email [email protected] with:

The affected URL or tool.
What you observed, and the steps to reproduce it.
Your browser and operating system.
Why you believe it has a security impact.

We typically review reports within 2-5 business days. Please give us a reasonable chance to fix a confirmed issue before disclosing it publicly, and please do not run automated scanners or load tests against the site or its CDN — that degrades the service for other people and tells us less than a clear reproduction does.

What is out of scope, stated honestly

This is not a bug bounty programme. There is no payment, no swag, and no formal reward for a report. We would rather say that up front than let you spend an evening on something expecting otherwise. What we can offer is that a valid report gets read, answered, and acted on.

Also out of scope: findings on domains we do not operate, issues in Google's products rather than ours, missing headers with no demonstrated impact, reports consisting only of automated scanner output, and the deliberate design decisions described on this page.

Abuse and rights concerns

Misuse of the tools is a different question from a vulnerability. The acceptable use policy sets out what these tools are and are not for, and explains the practical limits of enforcement given that we never see what anyone processes. Rights complaints go to the same address, and the contact page lists what to include.

Security & Safety