About AI Usage In Firewalla Support & Community

Announcements
1

Hi everyone,

We are starting to see a growing trend of support tickets and community posts that are largely written by AI.

We believe AI usage is great in certain applications, for example, sorting through repetitive alarms and understanding unknown domains or devices. However, when a ticket is AI-written, we often get very long descriptions of problems that don’t really match what actually happened.

We would love it if everyone could help keep the Firewalla Community more human-to-human:

  1. Tell us what happened in your own words.
  2. If you’d like to use AI to polish your words, please read it before posting or sending us a ticket.

We’ll also be updating our Contact Firewalla Support doc with a new AI policy:

  1. We prefer human-to-human interactions; please write to us with your own words.
  2. If you, for any reason, need to use AI to write, please keep it short, and please read it before sending it to us.
  3. If you send us a huge AI-composed ticket or response, we have the right to use AI to summarize it and may even use AI to respond.
  4. At the moment, the Firewalla Team does not use AI to respond to cases. (We may use macros, we may use AI to touch up the response, or change the tone, but never a technical response.)

If you have any suggestions for our new AI policy, please let us know. Thank you again for supporting Firewalla!

3 Likes
2

Also consider agentic ai that is trying to find ways of defeating or understanding feature setups in order to modify settings to disguise AI agent activity.

Sad to suggest this, but I can see an AI agent figuring out how to change rules, mute alarms, or allow access to specific ip and port combos if a firewalla box is compromised.

The above was neither checked or created by AI, although I did use my own biological LLM device. :smiling_face_with_sunglasses:

1 Like
3

That’s a reasonable concern. AI could potentially automate the process of understanding a configuration, changing rules, and hiding malicious activity. The critical prerequisite, though, is gaining enough access to make those changes. An AI agent’s presence on the network alone doesn’t give it that control.

If a firewall is already compromised at that level, a human attacker could attempt the same things. AI could make that process faster and more automated, which reinforces the importance of protecting management access, limiting agent permissions, and making administrative changes visible.

2 Likes
4

Potential avenue to instigate MFA for changes within the app or MSP side! Not just the app lock, but a ‘You are about to amend outbound rules, 0lease authenticate with your passkey’

1 Like