Just wanted to note a couple of things here.
First, there does appear to be an issue where the Answer with AI
toggle re-enables itself after being disabled. This is a bug and not intentional. I have opened an issue for this on our Github here:
opened 10:17PM - 20 Nov 24 UTC
feature/settings
OS/Desktop
Brave Search
browser-ai
### Description
When on search.brave.com, the `Answer with AI` toggle appears t… o re-enable itself after being disabled. This occurs sometimes even if the site is not refreshed (although it shouldn't happen regardless). Note that this occurs when you use the quick settings option on the Search homepage or when toggled off in the main settings page after clicking `All settings`:
https://github.com/user-attachments/assets/1cecb395-5a36-450e-b3b5-7070363e01c2
Many of our users do not wish to use AI for a variety of reasons and it is important that we provide a way to easily/reliably disable the feature and hide the icon for those users. Additionally, the option to answer with AI also appears in the search drop down field, which should also be hidden if the setting is toggled "off":

### Steps to reproduce
1. Visit serach.brave.com
2. Toggle `Answer with AI` option "off" in settings
3. Close settings
4. Reopen settings
5. Observe that the option has toggled itself back on
6. Additionally, observe that even when settings open and the option is toggled "off" the AI icon in the search field is still shown
### Actual result
Option toggles itself back on
### Expected result
Site settings should be maintained and this option should stay off and the related icon hidden.
### Reproduces how often
Easily reproduced
### Brave version (brave://version info)
v1.73.89
### Channel information
- [x] release (stable)
- [ ] beta
- [ ] nightly
### Reproducibility
- [ ] with Brave Shields disabled
- [ ] with Brave Rewards disabled
- [ ] in the latest version of Chrome
### Miscellaneous information
User reports of this:
https://community.brave.com/t/removal-of-answer-with-ai-from-search-box/544219/9
Second, on the note of AI giving incorrect answers, I just want to be clear that this is simply a property of AI/LLMs in general, not just Leo. This is referred to as the AI “hallucinating”. It is denoted in our Leo FAQ here and we encourage users to double-check/verify fact-based information when using [any] AI: