Added a button for an AI overview of the current website Terms and conditions to cover all websites (Optional , needs Open API key)

[jab442]

Screencast From 2025-06-29 12-01-05.webm

[jab442]

Thank you for your feedback , I respect your opinion and previous work on the project and I get where you are coming from.

But I see the AI button as a bonus feature- to even see the button you need to insert an API key in the settings. So the majority of users will not even know it is a feature. A better warning may be helpful, describing the potential limitations and potential for hallucinations.

I was wondering about allowing the system prompt to be changed of even customized in a textarea.The system prompt could receive instructions about the specific things you look for in the terms and conditions and seek them out and list anything suspect.

On Sun, 29 Jun 2025 at 12:55, shadowwwind @.***> wrote:

shadowwwind left a comment (tosdr/browser-extensions#114) https://github.com/tosdr/browser-extensions/pull/114#issuecomment-3016195073

We have a custom classification model that uses our existing systems and data to provide summaries. These summaries require human review. We also plan to eventually have clearly marked, temporary unreviewed summeries by the model. You can read about it here https://www.soundsandwords.io/privacy-policies/

ChatGPT is unreliable in this regard and imo reliability is important for our service.

— Reply to this email directly, view it on GitHub https://github.com/tosdr/browser-extensions/pull/114#issuecomment-3016195073, or unsubscribe https://github.com/notifications/unsubscribe-auth/AOGC37JICN2YOHDPTV62SMD3F42OPAVCNFSM6AAAAACALPO62GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTAMJWGE4TKMBXGM . You are receiving this because you authored the thread.Message ID: @.***>

[sonarqubecloud[bot]]

Quality Gate failed

Failed conditions
E Security Rating on New Code (required ≥ A)

See analysis details on SonarQube Cloud

Catch issues before they fail your Quality Gate with our IDE extension SonarQube for IDE

[ghost]

I would entirely disagree with such feature, I've spent quite some time using AI for such purposes, including this one specifically. And no matter what I used or which models I've tried, got regulary critical information wrong.

[ptgms]

Hey! Thank you for the PR, what you have made looks very impressive and we appreciate all contributions :)

AI and ToS;DR is a sensitive topic - we just started integrating a form of machine learning to ToS;DR and it is very delicate - it took us years to make a custom solution that satisfies our needs, does not hallucinate, etc. We believe trusting AI blindly can be deceiving, as it often makes up facts, points, and gets stuff wrong. The "AI" we use always requires human review.

What you have made looks really neat and is in concept cool, but I don't think it's a good idea to add this as it goes against our principals and guardrails we spent years working on.

Perhaps one day we make something like this when we feel it feels right, is private and most importantly, accurate.

Feel free to experiment with it further though, and if you have other ideas, you can reach out to us using our team mail ([email protected]). You seem like a very talented programmer and may be a great fit for our team :)