Jolla Unveils AI Assistant for Privacy-Focused GenAI

Jolla, once known for its mobile devices and now focusing on privacy through its sister startup Venho.ai, has just unveiled an AI assistant they claim is a "fully private" alternative to the big cloud companies that are always snooping on your data.

This AI assistant is set up to work seamlessly with your email, calendar, and social media apps, acting like a super-powered conversational tool. It can dig up info for you and even take actions like summarizing emails and documents, scheduling meetings, filtering your social media feeds, and doing web searches. But here's the cool part: it can also create new AI agents on the spot if it has the right API keys, expanding what it can do. They're planning to launch an AI agent marketplace next month to make this even easier.

They're also pushing the idea of using this AI assistant as a shopping helper and personal reminder system. You could get it to research products you're thinking about buying or have it jot down notes for you to remember later. And don't worry, all your personal bits and pieces stay private, not getting sucked into some giant data-mining machine in the cloud.

The AI assistant isn't just software; the "Jolla with Venho" crew has been working on special AI hardware for the past year. We've called it "AI agents in a box" before, and it's all about giving you personalized AI help without compromising your privacy.

They use smaller AI models that can run right on your device to handle most of your data needs, along with some preprocessing and a vector database to make everything run smoothly and quickly when you ask it questions.

The AI assistant software is designed to manage all this data juggling and switch between different AI agents as needed, making everything feel seamless.

Cutting through the noise

At the Mobile World Congress (MWC) in Barcelona, co-founder Antti Saarnio and Jolla veteran Sami Pienimäki gave TechCrunch a sneak peek at this new AI assistant tech. They think there's a big chance for startups to create a kind of decentralized AI operating system to shake up the big cloud companies, especially as generative AI keeps changing how we use software.

They're leaning on their experience with mobile hardware and OS development (think Sailfish) and integrating with Android apps, which they believe gives them a head start in this race.

"There's a huge amount of information coming at us because of AI, and we need help to sort through it all," says Saarnio. He thinks the AI assistant will be key to cutting through the clutter. "It's your own tool—you own that tool."

The AI assistant, branded as Mindy, comes with a customizable female avatar and is set to launch soon as a subscription service on a private cloud run by Venho.ai.

In a demo, Saarnio showed off the AI assistant on his laptop, typing in queries and getting responses in text and speech. He checked his emails, added tasks to a to-do list, and even booked a meeting. (You can also talk to it if you prefer.) There was a bit of a wait for the first response, but things got quicker after that. Saarnio mentioned they're working on making it even faster, aiming for a one-second response time.

A web search he tried didn't work because the laptop wasn't online, so the Google Search API couldn't do its thing. But no worries, they're sorting that out.

Besides offering the AI assistant on their private cloud, next month it'll also be available on the Mind2 Jolla AI device, which they've been sending out to early adopters since January. With Mind2, you can run the AI assistant right on your own hardware, keeping your data completely to yourself—even more private than the "private cloud" Venho offers. This works for queries that can be handled by the smaller AI models inside the device, like DeepSeek's 1.5-billion parameter model and Meta's Llama 1-billion parameter model.

"It can understand and give you accurate answers about your emails and their content without making big mistakes. But if you throw a huge document at it, it might get a bit confused," Saarnio explains, noting when the device might need to reach out to a bigger cloud for help.

For more complex queries, the tech might need to use larger language models (LLMs), which means your data could be exposed to someone else's terms of service. But with the conversational assistant in charge, you can tell it how to handle your data—like never sending health info to LLMs—making the experience more personalized and secure.

The early-bird price for the AI Assistant is $10 per month for the first 1,000 users (with a 14-day free trial), but it'll go up to around $20 per month. The Mind2 device, with 16GB RAM and 128GB memory, costs €699, though early adopters can still get a discount.

Since we last checked out Jolla's hardware, the Mind2 has gotten a bit bigger. Pienimäki says it's mostly because of the need for better heat management, with a larger heatsink and other components. The final assembly happens in Finland, at a former Nokia facility in Salo, which is pretty cool and might help bring back some of Finland's tech glory days. (Remember, Sailfish OS started as a fork of a Nokia project by former Nokia employees.)

So far, they've shipped around 500 Mind2 devices to early adopters, mostly from the Sailfish fan community. Saarnio says they're getting great feedback and help with bug fixes through a Discord community. "I'd recommend that to any startup," he adds.

While they're selling the hardware and services directly, they also see a big opportunity in the B2B market. Some telcos have spotted the potential for the hardware to act as a home hub for families, kind of like a "private Amazon Alexa," as Pienimäki puts it. This has pushed them to speed up the development of Mindy to show off what conversational AI can do.