I am highly sceptical of all AI features and it seems Gardner and other cyber security experts are starting to wake up as well:

The programs let you outsource and automate tasks, such as online searches or writing an email, to an AI agent. The only problem is that these same AI capabilities can be tricked into executing malicious commands hidden in websites or emails, effectively turning the browser against the user.

https://www.pcmag.com/news/security-experts-warn-companies-t...

For now this is restricted to Perplexity Comet and OpenAI Atlas and only the UK has issued an official warning, but why would I, personally, want my Firefox browser with an opt-out risk instead of an opt-in?