AI is everywhere. Our computers, phones and just about every SaaS offering has their own AI agent to "help us" get things done quickly, and with a level of quality that we often struggle to achieve on our own. I can't count how many times I've used ChatGPT to help make my writing clear, especially with web-copy or marketing materials. Despite the uncertain nature of where AI may take us in the future, it's here to stay.
While the advancement of this technology has been on a hockey-stick curve over the past few years, there's an important capability that is missing in today's solutions. AI solutions aren't personal. Ask ChatGPT, Gemini or any other publicly available AI agent anything about yourself and you'll get a response like this the below:
AI systems are trained using datasets. Very large datasets. ChatGPT and other public AI agents use data that is in the public domain - largely what's on the internet. So if you're a public figure, you may get some more personal results, but for the rest of us - it knows pretty much nothing about us. And that is a good thing - we don't want our personal photos, financial documents or our personal journals floating around on the internet for anyone to see - we keep those private, locked-away safely and under our control.
Even if we can feed our personal, private data to a public AI, that data is processed in cloud datacenters that power the AI engine which then begs the question - is that data still private to only you?
Personal AI Agents: Balancing Privacy and Performance
An ideal solution to this conundrum would be an AI agent, running locally on your computer and using only the data you provide for processing. Processing for your requests would be done locally, no internet connection needed, giving you truly private results. But, there's a problem with this, as well. Having a large trained dataset of public data is also needed, both for results but for the model itself. You see, training on your personal data isn't really a task you can easily do on consumer-grade computers. Instead, the trained datasets that are used for these public AI solutions can be downloaded locally to your computer to provide the rich breadth of knowledge required for any AI agent. Your personal data is then included through using a process called Retrieval-Augmented Generation, where the AI agent uses inference of your personal data in conjunction with the publicly trained data, now downloaded to your computer, to provide you with truly personal results like below:
Let's dive in a little deeper on what's happening here. In the image above, my knowledge base contains all of my photos, which subsequently has EXIF metadata containing the GPS coordinates of where the pictures were taken along with the date. Faces of individuals in the pictures have been tagged, along with their relationship to me (my family). Combine that with a pre-trained language model that can do object identification (like an elephant), as well as the friendly names of locations taken from the GPS coordinates of photos, dates, and my saved itinerary of the trip on my computer - and you get a complete response that provides context around facts I probably would have forgotten, like the name of the hostel we stayed at.
The key to success in having a truly personal and private AI is having a large personal knowledge base of information for the AI to reference. While the cloud is a convenient storage solution for our data, by not storing it somewhere under your complete control, you're giving-up some of your rights to that data (read your cloud storage agreement). So it's important that as much personal data as possible is stored locally, but then that begs the next question - what if I have too much data to fit on my computer?
What's needed to provide a complete solution is a data storage device with enough capacity to hold all of our personal data and the pre-trained AI models, but in addition adequate compute (CPU/GPU/NPU) and RAM to efficiently process the requests. A device powerful enough to not only answer questions, but also is capable of performing generative AI functions such as image and video creation. A device that you could access remotely, from anywhere, your own personal cloud AI agent.
DASSET Data Hub with AI
This is precisely what we have been working on at PlanetX Labs. With the introduction of the DASSET Data Hub in 2025, we are laying the groundwork that enables computer manufacturers (OEMs) to build devices that are capable of getting all of your personal data into a private repository that you control, which you can access from anywhere. Out of the gate, the DASSET Data Hub will be able to perform AI tasks such as photo face-tagging and scene classification. Intelligent search will enable natural language queries to easily find your data on your Data Hub. The note-taking app in DASSET will be able to summarize and generate content, making note-taking easy and intuitive. But this is just the tip of the iceberg.
Soon the ability to fetch, store and update trained language models will be built-in to DASSET, ensuring you have the latest trained models locally. Our AI agent, code-named "DAIA", will then use RAG to process complex natural-language queries with the ability to generate rich, truly personalized, responses not only like the example above, but much more that we're not quite ready to talk about just yet.
Personal AI needs to be accessible to anyone, not just those who are tech-savvy. Our mission at PlanetX Labs is simple.
...to give individuals, SMBs, and developers complete control over their data while allowing them to harness AI's potential.
We're excited for what's coming, and how it will enable everyday users to to interact with AI in a natural, human, way. The future's bright - and not that far away.