Jolla, the mobile pioneer turned privacy-focused AI developer, has officially introduced a new AI assistant designed to challenge data-mining cloud giants. Launched via its sister startup, Venho.ai, the tool promises a “fully private” alternative for users seeking to automate daily tasks without sacrificing personal data.
Conversational Power Meets Data Privacy
The assistant integrates directly with email, calendar, and social media accounts, acting as a conversational command center. Beyond simple information retrieval, the tool is capable of executing actions such as summarizing complex documents, managing meeting schedules, filtering social media feeds, and performing web look-ups.
Jolla plans to launch an AI agent marketplace next month, allowing users to deploy specialized agents on demand via API keys. The system is positioned as a personal aide, capable of researching purchases or storing notes for future recall, all while ensuring that sensitive snippets never leave the user’s private environment to feed corporate cloud empires.
Hardware-Backed Security: “AI Agents in a Box”
The software doesn’t operate in a vacuum. The team has spent the past year developing proprietary hardware, nicknamed “AI agents in a box,” to support this vision. By leveraging smaller, locally hosted AI models—including DeepSeek’s 1.5-billion parameter model and Meta’s Llama 1-billion parameter model—the system manages data tasks locally. A vector database unifies information across linked accounts, significantly boosting query speeds.
For users seeking maximum sovereignty, the assistant will be available on the Mind 2 hardware, which has been shipping to early adopters since January. This allows for self-hosting, effectively removing the need for third-party cloud interaction for most routine tasks.
The “Mindy” Experience
Branded as “Mindy,” the assistant features a customizable avatar and is currently rolling out as a subscription service. During a live demonstration at Mobile World Congress (MWC) in Barcelona, co-founders Antti Saarnio and Sami Pienimäki showcased the assistant handling email queries and meeting scheduling via both text and voice commands.
While the team is currently optimizing response latency—targeting a one-second turnaround—the system architecture allows for a tiered privacy approach. For highly complex queries requiring massive LLMs, the system allows users to define strict boundaries, such as preventing health data from ever being transmitted to external cloud models.
Strategic Growth and Market Positioning
The Mind 2 hardware has evolved in design, now featuring a robust heatsink to manage the thermal requirements of continuous on-device processing. Assembly takes place in Salo, Finland, at a facility formerly owned by Nokia, signaling a potential revival of Finnish hardware expertise.
With roughly 500 units already in the hands of enthusiasts, the feedback loop remains tight. The company is now eyeing B2B opportunities, particularly with telecommunications providers interested in “home hub” solutions. As Saarnio notes, the goal is to provide a “private Amazon Alexa” experience, helping users navigate the overwhelming cognitive load of the modern AI era while maintaining complete ownership of their digital footprint.