The Local-First Promise: Auditing Synsira's On-Device Claims
AI-generated image · Bay Street Wire
A new on-device AI tool promises total data sovereignty, but the devil is always in the telemetry.
In a landscape dominated by cloud-sucking LLMs, the only acceptable threat model is one where the data never leaves the silicon. Most 'privacy' claims are just marketing veneers for slightly more efficient data harvesting. That is why the launch of Kind Local Pro (KLP) from Synsira Software caught my eye.
According to reporting from BetaKit, Jonathan Schaeffer—a former founder of the Alberta Machine Intelligence Institute (Amii) and current head of the Vancouver Island-based Synsira Software—has released a version of his flagship product designed to run entirely on-device. While the standard Kind Pro utilizes cloud infrastructure, KLP is marketed as a privacy-first alternative that allows users to interact with AI without internet access.
Schaeffer told BetaKit that all AI processing occurs on the user's computer and that data never leaves the machine. The tool leverages open-source AI models and allows users to ingest a variety of files, including email inboxes, audio, images, documents, and video, to create a localized knowledge database. Schaeffer explicitly contrasted KLP with Google's Notebook LM, stating that KLP does not move user files to cloud storage.
From a defender's mindset, the critical question is always: *What is the actual leak path?*
Schaeffer admitted to BetaKit that there is one point of contact between the installation and Synsira: periodic account verification to confirm a valid license. In the world of cybersecurity, 'periodic verification' is where the telemetry often hides. Until we see the packet captures, the claim of being 'local' remains a promise, not a proven fact.
Beyond the privacy angle, BetaKit reports that Synsira is targeting data-sensitive sectors such as academia, intellectual property development, and legal work. To mitigate safety risks associated with a lack of cloud oversight, Schaeffer noted that KLP does not offer 'chatting' and avoids anthropomorphizing interactions, instead providing fact-based answers derived solely from the user's provided data.
Schaeffer also framed the shift as an environmental necessity. He told BetaKit that by reducing reliance on massive, expensive data centers, on-device processing can lower the overall environmental impact of AI.
**Opinion:** Local-first is the only way forward, but 'local' is a spectrum. If a tool requires a phone-home for license verification, it is not air-gapped. I will be watching to see if Synsira's 'ethical AI' is truly a fortress or just a slightly more polite way to manage user data.