Former Amii founder launches privacy-first AI that keeps user data local

A former Alberta Machine Intelligence Institute (Amii) founder, who has since gone on to start an “ethical AI” company, has released a privacy-first model intended to keep all data in the hands of the user. 

 
“I’m unhappy with the way that AI is going these days, and I wanted to do something about it.”

Jonathan Schaeffer,
Synsira Software

Synsira Software is a Vancouver Island-based AI company founded by Jonathan Schaeffer, one of the AI researchers who helped found Amii. Synsira says it builds “ethical, user-friendly AI products” with a privacy and environmental impact focus.

The firm’s latest product offering is a version of its flagship product, Kind Pro. Whereas Kind Pro sends portions of user data to cloud infrastructure, this new version runs entirely on-device, enabling users to interact with the AI program without accessing the internet, as well as retain control over the data those interactions generate. 

Dubbed Kind Local Pro (KLP), Synsira’s product installs proprietary AI tools that leverage open-source AI models. Once installed, users drag-and-drop files—ranging from documents, video, images, videos, email inboxes, audio, and more—into KLP. In turn, the AI ingests those documents and bases its knowledge database and responses solely from the materials provided. 

While similar in concept to applications like Google’s Notebook LM, KLP differs by not moving any user files off-device to cloud storage or elsewhere, according to Schaeffer. 

“All AI processing happens on your computer; your data never leaves your computer,” Schaeffer told BetaKit in an email. “Once you install Kind, there is only one interaction between your installation and Synsira—periodic account verification to ensure you have a valid license.” 

Schaeffer, who has been working on Kind to different degrees since 2017, said he was motivated to develop privacy-oriented AI software after becoming disillusioned with the way many large AI companies were approaching data privacy. 

“With the advent of LLMs—in particular, ChatGPT … in 2022—I realized how important Kind was for helping respect the privacy of people’s data,” he said. “To be candid, I’m unhappy with the way that AI is going these days, and I wanted to do something about it.” 

Because KLP operates independent of cloud-based storage and data centres, Synsira is marketing the product toward data-sensitive industries like legal work, intellectual property development, or academia. 

Balancing act

KLP’s hands-off approach means more data agency for users, but it also presents potential vulnerabilities. Recently, many of the world’s largest AI companies have come under pressure to do more to curb potentially harmful interactions between users and AI platforms, and the content they generate. In Canada, in the aftermath of the Tumbler Ridge, BC mass shooting, ChatGPT parent company OpenAI was pressured by the federal government to put in place stronger guardrails that would alert authorities if a user’s interactions set off red flags around self-harm or the potential for violence. 

With KLP running locally, and without any outside oversight of how the platform is being used, BetaKit asked Schaeffer how Synsira was balancing the right to privacy with calls for safety.  

RELATED: Alberta to invest $50 million into Amii over five years

“We have guardrails in place to prevent users from manipulating KLP into an inappropriate answer. Kind products do not offer “chatting”—we view this as a safety risk. We do not anthropomorphize interactions,” Schaeffer said. “Second, all queries to Kind result in only fact-based answers, where facts come from only your data.” 

Schaeffer said he views Kind’s interface as being akin to the computer in Star Trek. 

“The computer has no personality. You ask a question and the computer does the analysis to come back with a fact-based answer. When it cannot get an answer, the computer lets you know that,” he said.

Looking toward a different status quo

Because KLP is designed to run in the resource-restricted environment of a typical desktop computer, and because it intentionally avoids the use of data centres and cloud infrastructure for storage, Synsira’s platform carries with it smaller energy demands.  

“The environmental movement wants people to think globally, but act locally. That is part of the motivation for KLP,” Schaeffer said, adding that if more AI processing is done on-device, the need for data centres comes down. 

“We can have a global impact by reducing the demand for massive, expensive, environmentally unfriendly data centres,” he said.

BetaKit’s Prairies reporting is funded in part by YEGAF, a not-for-profit dedicated to amplifying business stories in Alberta.

Feature image courtesy of Synsira Software