I am creating my own AI assistant who reads all my emails, messages, and all new files every single day. It gives me a daily rundown of what I have knocked off my to-do list and what’s coming. I even had it process my entire historical record so I have a searchable database of all my thoughts and communications. I run it on Scaleway, on a rented GPU.
It boots itself up, takes in yesterday’s feedback to improve itself, and then it kills itself. Every day.
Kill is a strong word. It shuts itself down, “gracefully” as the polite term. I have an image (basically, a digital “snapshot” of the technical setup) saved so when I restart it I save time and effort.
I built it this way because there was nothing on the market I felt good handing my data over to. As someone who both loves AI and is deeply concerned by the way it has been deployed on the commercial market, I want to play with AI, but I don’t want to get burned.
The trends in privacy are starting to reverse themselves#
For the past two decades, we have consistently traded in our privacy for convenience. Many of us, especially people who would call themselves socially progressive, believed the world was getting better: that if you had nothing to hide, you had nothing to fear. Surveillance wasn’t scary to a lot of normal people living seemingly normal lives which may have included LGBT advocacy, sharing details of their experiences as a disabled person, or even calling oneself a feminist, as many female tech workers would identify.
However, it’s become more dangerous to be yourself. Governments, especially in the USA and Britain, are interested in invading people’s privacy in order to know who to track, and while your beliefs might not make you a candidate today, they could in the future. Five years ago, the US government was funding all sorts of LGBT initiatives. These days, they’re using Grok to remove their funding. They’ve also stopped recognizing certain religions in their military, leaving questions unanswered about religious accommodations for people who fall outside the new definitions.
That’s why I see the value of our privacy exponentially increasing — and the bubble risk will be for all those who invest in invading that privacy. Encouraging us to part with our data privacy over decades was a long-term investment by tech companies to find new things to commoditize, but what happens when the well dries? If you think that is far-fetched, I invite you to ponder…
What happens when all of a sudden, the knowledge on how to protect your data privacy is easily accessible via the same tools created to monitor you?
GDPR was a small nail in the coffin for the tech giants who wanted to know every detail of our lives, but I feel they’ve added a few more themselves by ingesting the entirety of human knowledge and selling it back to us for free-99. Not only that, but they’re not seeing the trends — that in places like the EU, where collective action is still able to move the needle, there is a strong desire for data privacy. That maybe people who were already working on data privacy projects might take advantage of the extra help AI provides to further accelerate access to personal and organizational data privacy. That maybe there are people in the open source world who want to give access to high-quality models in a way that respects data privacy.
France, as I’ve mentioned before, is moving away from foreign technology for their civil servants. Here in Luxembourg, we are the first country to have signed a strategic partnership with an EU-sovereign frontier model, Mistral. It’s not exactly how I would have done it, but it’s better than ChatGPT. Individuals like myself are building their own private AI tools, and even Gen Z has managed to add their charming appreciation for aesthetics to this trend by making cyberdecks — personal, homemade computers that can be as locked down as you wish.
Now, this bubble is far from being popped. Individuals, companies, and even countries continue to hand their data over to companies like Palantir, Google, Amazon, Oracle, and Meta, to name a few. Don’t even get me started on “X the everything app.” And AI has, of course, also exacerbated this. It seems like a lot of people don’t really think twice about giving intimate details of their lives to computers — maybe we’ve been conditioned to this through social media.
But sometimes people are tricked: a grandmother submitting her grandchild’s likeness to a company on the App Store to make a custom storybook for the child’s birthday. An influencer telling women to use AI for their financial advice, making their private information vulnerable, and possibly resulting in financial harm. Neighbourhoods installing Flock cameras for their own protection, only to realize they now live under constant surveillance, and not always by people they can trust. This isn’t just an American thing — it’s also being rolled out in Germany.
People are starting to push back. You’re seeing more and more discussion about privacy issues, with some people taking the no-AI approach (laudable, quite frankly) and others, like me, trying to see if we can steer this ship better. Some people think it’s futile, but I am nothing if not an optimist.
The big change now is how easy it is to create your own sovereign tech stack. The next evolution is going to be people running their own businesses with minimal cost for SaaS subscriptions, keeping their data safe, and even using old technology. My private AI assistant runs via a daily cron from my 11-year-old MacBook. It simply turns on my rented GPU every day for a cost of about €0.20 until I feel like investing in a setup that can run it locally.
Where does this leave the frontier models, if there’s a path to getting off of them?
If you’re interested in setting up private AI assistance for your home or your company, feel free to get in touch.