$700 million. That’s what Hark raised in a Series A, at a $6 billion post-money valuation, for a company founded less than a year ago. For context, that’s a funding round larger than most startups see across their entire lifetime, handed to a team whose product was still mostly a rumor.
Now the product is real. Hark has released Hark Pro, an AI personal assistant built around a specific promise: your data stays on your machine.
What “privacy-first” actually means here
If you’ve used an AI assistant before, you’ve almost certainly used one that works like a phone call. You type or speak, your words travel to a data center somewhere, a large model processes them, and an answer comes back. That round trip is why these tools feel fast and capable. It’s also why privacy questions keep coming up. Your calendar, your emails, your half-finished documents, your messy personal notes — all of it has to leave your computer for the assistant to be useful.
Hark Pro is designed to run on your own computer instead. The assistant uses your machine’s resources to do its work, which means the sensitive material it touches doesn’t need to be shipped off anywhere. That’s the pitch, and it’s a meaningfully different architecture from the default.
For non-technical readers, a useful mental model: most AI assistants are like hiring a brilliant consultant who insists on taking your files back to their office. A local assistant is more like one who works at your kitchen table and never takes anything with them.
Why this is landing in 2026
Demand for tools where data doesn’t leave the user’s machine has been building steadily. People have gotten more specific about what they’re worried about. It’s less abstract dread about surveillance and more practical: I don’t want my work-in-progress, my client contracts, or my medical questions sitting in someone else’s logs.
Meanwhile, the industry has been in an open hunt for the user experience that makes an AI assistant something you actually can’t live without. TechCrunch framed Hark’s release alongside Muse, Instinct, and Dots — a cluster of companies all circling the same question from different directions. Nobody has definitively won. Chatbots were the first answer. They were fine. They were not indispensable.
Privacy could be the differentiator that tips an assistant from “neat” to “necessary,” particularly for professionals handling information they’re contractually or legally obligated to protect. Lawyers, therapists, accountants, and anyone under an NDA have a concrete reason to prefer local processing that has nothing to do with ideology.
The bigger plan looks past the assistant
Here’s the part worth paying attention to. Hark was founded by Brett Adcock, a serial founder, and the company’s stated goal isn’t just to ship a good assistant. It’s to build an operating system for AI computers of the future.
That reframes Hark Pro considerably. The assistant isn’t the destination; it’s the first visible piece of a much larger bet. Hark has promised an AI-native device in 2027, and its funding describes proprietary multi-modal models alongside custom hardware.
Put those pieces together and the strategy reads something like this:
- Build models you own rather than renting someone else’s
- Ship software that runs locally, proving the approach works on existing computers
- Design hardware built specifically for that software
- Position the whole stack as an operating system, not an app
Companies that control both the software and the hardware can make choices others can’t. Running AI locally is demanding — it needs processing power and memory that general-purpose laptops weren’t designed around. Custom hardware is one way to solve that, and it explains why a privacy-focused assistant and a 2027 device announcement belong in the same story.
What to watch for if you’re considering it
Local-first assistants come with real tradeoffs, and anyone evaluating one should go in clear-eyed. The questions I’d ask of any tool in this category:
- How much of the work genuinely happens on your machine, and does anything still get sent out?
- What does it need from your computer to run well?
- What permissions does it want, and can you limit them per app or per folder?
- If you leave, what happens to the data and models stored locally?
Those questions matter more than any benchmark. Permission controls and clear data governance are what separate an assistant you can trust with real work from one you keep at arm’s length.
A bet with a deadline
Hark has a lot of money, a short history, and a very specific thesis: the next generation of personal computing will be organized around AI, and the winner will be whoever owns that layer end to end. Hark Pro is the opening argument. The 2027 device is where the argument gets tested.
For the rest of us, the useful takeaway is simpler. Privacy has moved from a feature bullet to a design constraint that companies are willing to raise hundreds of millions of dollars against. That shift benefits you whether or not Hark ends up being the one to pull it off.
🕒 Published: