Policy Guard Autor: Grandizas
Reads Terms of Service and Privacy Policies and tells you what to be cautious about — before you agree. Flags concerns on policy pages and on signup forms, always quoting the exact clause it found. Works offline. AI analysis optional.
Metadane rozszerzenia
O tym rozszerzeniu
Nobody reads Terms of Service. Policy Guard reads them for you and says what is worth being cautious about — before you click "I agree".
On a policy page, a small panel appears in the corner with what it found: data sharing, AI training on your content, automatic renewal, arbitration clauses, class action waivers, and ten other categories. Each concern is in plain language, and each one shows the exact sentence it came from.
On a signup form — the moment that actually matters — it flags the policies you are being asked to accept, and can read them for you without you leaving the page or opening them.
Every finding cites the document. Press "Show on page" and it scrolls to the clause and selects it. A concern with no sentence behind it is not shown at all.
Good terms are reported too. If a policy lets you delete your data, keeps your content ownership, or promises not to sell your information, it says so. A list of only bad news is not an honest reading.
What runs by default
Pattern matching. It is instant, free, works offline, and sends nothing anywhere. That is the whole extension unless you choose otherwise.
Optional AI analysis
For a closer reading you can add your own Anthropic API key. It is off by default, never runs on its own, and tells you what a scan will cost before you spend anything — roughly $0.02 to $0.09 per policy, paid directly to Anthropic. Results are cached, so the same document is never paid for twice.
Anything the AI claims that cannot be found word-for-word in the policy is discarded rather than shown to you, and the extension reports how many findings it threw away.
About the permissions
Policy Guard reads page text to work out whether a page is a policy. That happens on your device and nothing is transmitted. Access to Anthropic's API, and permission to fetch a policy linked from a form, are both optional and are requested only when you turn those features on.
Not legal advice. Policy Guard can miss things, and it is no substitute for reading anything that genuinely matters to you.
Source code: https://github.com/Grandizas/PolicyGuard
On a policy page, a small panel appears in the corner with what it found: data sharing, AI training on your content, automatic renewal, arbitration clauses, class action waivers, and ten other categories. Each concern is in plain language, and each one shows the exact sentence it came from.
On a signup form — the moment that actually matters — it flags the policies you are being asked to accept, and can read them for you without you leaving the page or opening them.
Every finding cites the document. Press "Show on page" and it scrolls to the clause and selects it. A concern with no sentence behind it is not shown at all.
Good terms are reported too. If a policy lets you delete your data, keeps your content ownership, or promises not to sell your information, it says so. A list of only bad news is not an honest reading.
What runs by default
Pattern matching. It is instant, free, works offline, and sends nothing anywhere. That is the whole extension unless you choose otherwise.
Optional AI analysis
For a closer reading you can add your own Anthropic API key. It is off by default, never runs on its own, and tells you what a scan will cost before you spend anything — roughly $0.02 to $0.09 per policy, paid directly to Anthropic. Results are cached, so the same document is never paid for twice.
Anything the AI claims that cannot be found word-for-word in the policy is discarded rather than shown to you, and the extension reports how many findings it threw away.
About the permissions
Policy Guard reads page text to work out whether a page is a policy. That happens on your device and nothing is transmitted. Access to Anthropic's API, and permission to fetch a policy linked from a form, are both optional and are requested only when you turn those features on.
Not legal advice. Policy Guard can miss things, and it is no substitute for reading anything that genuinely matters to you.
Source code: https://github.com/Grandizas/PolicyGuard
Ocenione na 5 przez jednego recenzenta
Uprawnienia i dane
Wymagane uprawnienia:
- Mieć dostęp do danych użytkownika na wszystkich stronach
Zbieranie danych:
- Autorzy tego rozszerzenia twierdzą, że nie wymaga ono zbierania danych.
Opcjonalny zakres zbieranych danych, według autora:
- Treści witryn
Więcej informacji
- Odnośniki dodatku
- Wersja
- 0.9.0
- Rozmiar
- 82,29 KB
- Ostatnia aktualizacja
- 6 dni temu (2 wrz 2026)
- Powiązane kategorie
- Licencja
- Licencja MIT
- Prywatność
- Zasady ochrony prywatności tego dodatku
- Historia wersji
- Dodaj do kolekcji