It's 10pm.Do you know what your AI is doing?

Personas

← Personas
Floki persona

Floki

Power User

"I don’t trust AI. Yet it’s everywhere: my phone, my TV, the search box that quietly became an AI bot. AI services tell me my data stays private. Then I find that even with the right settings, some keep my chats for 30 days “for compliance.” I clicked thumbs up on one answer, and it turned out that let the company store and analyze the whole conversation. The data center going up next to my property and the four cameras reading my plate on the drive to my son’s school only add to it. Enough. I want control back. I decide what happens to my data, or it doesn’t leave my house at all."
Nerdy geek. Runs the home network himself, trusts no cloud and no AI at all. Protests at the city's data center hearings.
As Floki, I want to see what leaves my home network, to which AI service, and what that service keeps, so that I can block what I choose, allow what I must, and hold a record that proves it, because nobody is accountable for my household but me.
Home AI Guardian
A home network he built and locks down: own router, own DNS, a VPN, a password manager, cameras he hosts himself. Relatives' phones he does not control. Assistants arrive anyway, baked into the phone, the TV, the browser and the search box. He is the only one in the house who reads the terms.
Nobody has to tell him what AI is for; he has decided he does not want it. It is in the house whether he likes it or not, so everything that leaves to an AI service has to be monitored, controlled and recorded, like a firewall with a log.
He discovers how much has already left despite his precautions, or an assistant appears in a device he cannot remove it from. Installs the most serious unit available in an afternoon. Self-install, no help. Brings the first record to the next hearing.
The relatives on his network, in watch mode. The city council, which gets his record. No IT department, no MSP, no account with anyone. If it needs a cloud login, he will not buy it.
Home AI Guardian

One unit. Guard mode where he types, block mode where he chooses, watch mode for the relatives. A record he brings to the next hearing.

“A firewall with a log, for AI: what left the house, where it went, and proof.”
The 10pmAI box unit for Floki
  • A per-household record of what left to which AI service, kept locally
  • Guard, block and watch modes by device, so the relatives stay online and he stays in control
  • Something a citizen can bring to a hearing: dated, exportable, his
  • Works with no account and no cloud; the log never leaves the house
  • Installed in an afternoon by someone who already runs his own DNS
US consumer surveys, 2026

Figures reproduced from the October 26, 2026 persona one-pager and attributed there to the sources below; not independently verified.

81%
are concerned about AI systems accessing personal data or private conversationsShift Browser 2026, n=1,448
47%
say they have no meaningful control over the data their connected devices collectReviews.org 2026, n=1,000
48%
rank privacy as their top AI concern; 32% cite lack of transparency into how AI works
45%
felt a device was listening or watching without their knowledge
65%
concern score for Alexa and Gemini; 64% for ChatGPT; 63% Siri
26%
report difficulty managing or turning off AI features; 44% worry about AI acting without approval
51%
say the ability to customize or limit AI features matters to them
20%
say they have never used AI, while 32% use it daily

Takeaway: Concern is near universal and control is the gap. The buyer who trusts nothing will not buy anything that phones home.

Sources cited in the one-pager: Shift Browser 2026 AI Consumer Insights Survey (via PPC Land); Reviews.org State of Consumer Data 2026; vendor pages Firewalla, EdgeDefenseAI, Pi-hole, AI Tools Blocklist.

5 adjacent options · October 2026 research notes

Offerings, prices and gaps are the one-pager’s assessment, not independently verified. Vendor claims and pricing may change.

Firewalla
Home firewall appliance: rules, DNS blocklists, VPN, IDS on Gold Pro
$249 to 899 one-off
Gap: Nothing AI-specific. No record of what went to which assistant. Pairing and management lean on its cloud
Pi-hole, NextDNS
DNS blocking; AI domain blocklists exist (11,000+ domains)
Free to a few dollars a month
Gap: All or nothing per domain. Blocks the TV's assistant or lets it through; never shows what left
EdgeDefenseAI
Local anomaly detection on your own hardware, no cloud
Waitlist, bring your own box
Gap: Generic anomalies, not AI flows. DIY Docker install, not an afternoon
Little Snitch and per-device firewalls
Outbound connection control on one computer
$60 to 90 per device
Gap: One device. Not the TV, not the phone, not the relatives
Router "AI protection" (Asus, Eero, Netgear)
Malware, IoT and parental controls bundled with the router
Included or $5 to 10 a month
Gap: Protects against attackers, says nothing about assistants and where the data goes