Ubroker Locally: Erase Yourself From 500+ Data Brokers For Free

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Ubroker Locally: Erase Yourself From 500+ Data Brokers For Free · Channel: Fahd Mirza · 08:18 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Open source and software ecosystems, Privacy and data protection

Summary

The video introduces Unbroker, an open-source Hermes agent skill that automates much of the process of finding and opting out of data brokers while running locally. The presenter installs and configures it, expands its broker list, and demonstrates the workflow with test data before trying it with his own information. He finds that the tool is focused on the US and Canada and does not locate his information in Australia. He also cautions that it is early-stage, uses plain-text data by default, and still requires human help for some opt-outs.

Key points

Data brokers collect and sell personal information, which can contribute to stalking, doxing, and identity theft.
Although privacy laws provide deletion rights in some regions, each broker has a different opt-out process.
Unbroker is a free, open-source Hermes skill designed to run locally and automate broker scans and opt-out requests.
The skill builds identity profiles that include details such as previous addresses and email addresses.
The demonstrated setup uses draft-only emails and stores its dossier as plain-text JSON unless users configure additional protections.
The presenter expands the initial list of 22 brokers to roughly 50-plus using additional lists and a California registry.
In the test workflow, the tool recommends a California deletion request and scans across people-search sites.
The tool is primarily US-focused, and adding brokers from other regions can be cumbersome.
Hard CAPTCHAs and requests requiring calls, government ID, or fax are left for a human to handle.
The presenter describes Unbroker as an early-stage project rather than a mature, fully automated deletion service.

Chapters

Start	Chapter	Summary
00:01	The data broker problem and Unbroker	The presenter explains how brokers expose personal information and introduces Unbroker as a local, free tool for automating opt-outs.
02:05	Installing and configuring the skill	He explains Hermes skills and installs Unbroker, then runs its configuration and readiness check.
04:10	Broker lists and a test run	The presenter adds broker lists, creates a fake test subject, and reviews the tool’s suggested next steps.
05:44	Regional limits and safety drawbacks	A run using the presenter’s information finds no results, and he discusses the tool’s geographic and operational limitations.
07:16	Early-stage status and conclusion	He notes that some tasks still require human action, characterizes the tool as early-stage, and encourages viewers to consider trying or adapting it.

Quotes

Your data never has to leave your machine. — 01:02

Full autonomy, but draft only emails, no encryption, no cloud browser — 03:39

it is only US and Canada focused. — 06:15

This is not a magic 100% bulk delete button. — 07:16

you are actually testing it alongside the developer — 07:46

Concepts and entities

Unbroker, Hermes, Nous Research, DeleteMe, Bad Bull, OpenClaw, NeoClaw, Fahd Mirza, CCPA, GDPR

Full transcript (original language)

00:01 Right now, if I wanted to, I could find out where you live in about 10 seconds. Not because I am a hacker, because hundreds of data broker sites have already scraped your information, even your email, phone number, your relatives, and they are selling it just to publish it outright. This is the exact same data that fuels stalking, doxing, and identity theft, and of course, the AI model training.

00:31 And yes, laws like CCPA in California and GDPR in Europe say you have the right to make them delete it. But there is no bulk delete button. Every single broker has a different opt-out process, and a lot of them make it deliberately painful. That gap gap is the entire business model behind various services like DeleteMe and various other than they charge you quite handsomely. And ironically, you’re handing your

01:02 exact sensitive data to yet another company in the process. That’s exactly the gap this skill unbroker closes. It’s a new skill for Herme’s agent open-sourced by its creator that does the entire workflow for you for free. Running locally on your own machine, it builds out every version of your identity, current name, old addresses, aliases, past emails, because brokers often index

01:32 you under stale information. Then it fans out and scans the broker list in parallel, files the actual opt-out requests, handles the CAPTCHA and email verification loop where it legitimately can, and only comes back to you with a short list of things that truly need a human, like a phone call or a government ID upload. Your data never has to leave your machine. If you don’t know what Hermes agent is, it is a Now Research open self-improving AI agent that runs from your terminal.

02:05 It can call tools, connect to messaging platforms, and extend itself through installable skills. And we have covered it a lot on the channel. Um now, this tool, Unbroker, is not perfect. There are various limitations and drawbacks. I will mention them as we go along the video. But for now, let’s get it installed, and then we will go from there. And by the way, this is Fahad Mirza, and I welcome you to the channel. Recently, I also have started this free weekly AI newsletter, which you can subscribe at

02:35 fahadmirza.substack.com. I’m going to use this Ubuntu system. I already have Hermes installed. If you have it installed, just make sure to update it as this skill is only available in the latest version. And if you don’t know what skill is, a skill is basically a packaged instruction set, a skill.md file, plus supporting scripts that teaches Hermes exactly how to carry out a specific repeatable task, so the agent doesn’t have to improvise it from

03:06 scratch every time. So, I’m just going to install this skill easily. And that’s all done very, very quickly, as you can see. And there are various other skills quite useful. And by the way, not in Hermes not only in Hermes agent, but in lot of other code harnesses like OpenClaw and NeoClaw, and the list goes on and on and on. After installation, I have gone into the directory where that skill got installed in dot Hermes folder. And these commands

03:39 are configuring itself, this skill, and run a readiness check before we actually point it at anyone’s data. So, let me run this. And it was fairly quick. You can see that it auto configured itself into the safest possible mode. Full autonomy, but draft only emails, no encryption, no cloud browser, and confirmed it’s ready with 22 curated broken loaded. Um >> [clears throat] >> Now, after this, let me now pull the

04:10 five live broker list from Bad Bull plus the CA data broker registry, expanding coverage from 22 brokers to roughly 50 plus. And of course, it is all code. You can also grab it from where wherever you like. You see that I have also uh pulled all of these data brokers where your information might be. So, now we have got the full picture. Let’s actually create a test subject with fake data and consent since nothing

04:42 runs without data. You see that I have just put in some test person and city and few other things. Let’s run this. And that command has created the test subject with this ID. Confirmed authorized California residency, one email, and one address on a file. And this subject ID is now how every future command references this person. So, now let’s ask it what to actually do next. This is autonomous loop driver, which I’m running at the moment. Let me

05:13 take you up to show you the output. So, if you look at this output and especially focus on this counts, you see it has found nothing because it was just a test person. That’s expected. It hasn’t scanned anything. It’s just telling you the two moves to make next. Since you’re since this person is in California, file the one CA drop request that wipes you from all five 24 registered brokers at once. And separately kick off a parallel scan across all 52 people

05:44 search sites to see which ones actually have you listed. And then you can simply go and check it out. Now, I’m not sure if the residency is Australian or not, but nonetheless, I’m just going to go and run this because I think so far it is just only in Canada and US, but let me run it with my name and my city. And you see that it has given me this response. And at least hasn’t given me any error for Australian residency. Anyways, so

06:15 I’m just going to run this and see what it does. So, as was expected, it didn’t really find any information regarding me because it is only US and Canada focused. You would have to change it in the code. Uh, and it is quite cumbersome to add a data broker from your region. [clears throat] I think this is one improvement they could make to make it globally accepted because there is a world outside of California. Now, not only that, there are also various other limitations of this skill.

06:45 And that is a problem with these agents that if we cut out the hype, um, first as I said, it’s US-based. The broker database, the bad boot list, and the entire registry land are built around American data brokers. Also, another, um, thing is that, you know, it’s default zero config state. It runs in draft only mode with a dossier stored as plain text JSON on disk. No encryption, no auto send emails until you deliberately 100 API keys or

07:16 install age. So, hands-off isn’t actually the out-of-box behavior. You have to opt into that. Third, it does not defeat hard captchas or anti-bot systems and anything requiring a phone call, government ID, or fax. Um, that’s get pointed straight to a human task digest. So, this is not a magic 100% bulk delete button. It’s a well-structured assistant for a genuinely tedious, high-friction chore. And finally, this is very early stages shipping

07:46 a lot of placeholder data. So, I think if you’re using it right now at this early, you are actually testing it alongside the developer, not using a mature battle-hardened product. But look, I still believe worth trying to worth understanding and worth using. And if you’re looking to have something like that in your project, maybe just fork it and convert it into your own use case. That’s it. Let me know what do you think about this tool. Again, if you want to help out the channel, please

08:18 become a member. Thank you for all the support.