M12 Ventures, the venture capital arm of Microsoft, and astronaut Chris Hadfield are among the diverse investors in Private AI, a developer of privacy-preserving machine learning and natural language processing tools.
Along with Forum Ventures, M12 co-led a $3.15 million round of seed funding for the Toronto-based AI firm. Pre-seed investor Differential Ventures also participated in the heavily oversubscribed seed round.
Private AI’s tool is used to redact sensitive information from text.
The company also attracted new investors Shasta Ventures, Hyperplane Venture Capital, and Parliament Angels, a group of early Twilio employees, and Ajay Agrawa, the founder of Creative Destruction Lab.
The funding round closed in early September.
Private AI’s tool is used to redact sensitive information from text. The company’s AI models are able to achieve greater than 99 percent accuracy in identifying and redacting personal data across more than fifty different entities, such as name, address, blood type, zodiac sign, credit card number, in seven different languages.
The AI system performs particularly well on messy, real-world text, such as emails, chat messages and free-text fields in databases.
“If you have a model like a chatbot that’s spewing out information about an individual that it’s memorized the training data from the previous chats they’ve had, that could definitely be a threat,” Patricia Thaine, Private AI’s co-founder and CEO told BetaKit.
The tool is also used to protect the privacy of individuals when it comes to things like the General Data Protection Regulation, the EU’s privacy and security law.
Private AI plans to use its latest funding to double the size of its team from 10 to 20 by 2022, and to expand its current technology offering.
The round followed a sixfold revenue increase since January. Private AI is a private company, and declined to disclose its financials. The firm’s customers now range from startups to multi-billion-dollar companies, including financial institutions and conversational AI companies.
Asked about Hadfield’s interest, Thaine said, “I think he thinks the product is really cool.”
Investors from M12, Forum, and Differential will also join Private AI’s board as observers as part of the deal.
“Our goal is to create a ‘Twilio for privacy.’ So, it’s very significant for us to have both their product and technical help,” Thaine said, adding that it would involve “taking a huge amount of complexity and making it available in only a few lines of code so that any developer can easily integrate privacy into their software.”
San Francisco-based Twilio is a cloud communications platform. The investment from Parliament Angels, a group of early Twilio employees, will strengthen the ties between Private AI and Twilio, a company Private AI hopes to emulate.
Private AI began in 2017 with a different idea related to natural language processing. The original idea was to enable searches of encrypted data, but without having to type in every keyword. However, Thaine said that even though a couple of financial institutions were interested in the idea, it wasn’t going to scale to the degree Private AI wanted.
In 2018, the company returned to research, and in 2019 “we really started to try to figure out what the market needed, and that’s where we saw this massive gap where developers had no really easy way of integrating privacy into their software,” Thaine said. “That’s what we set out to solve.”
Along with Thaine, Private AI was started by Pieter Luitjens (CTO), and Gerald Penn (chief science officer). Thaine and Luitjens hold Master’s degrees in computer science and engineering, respectively, from the University of Toronto. Penn is also a professor of computer science at the University of Toronto.
The firm previously received $300,000 in pre-seed funding in June 2020 from Differential and Forum (then known as Acceleprise).
Private AI faces some large competitors in the space, which includes Amazon and Google. Both companies have cloud offerings for detecting personal data in your information. However, Thaine pointed out that both send peoples’ data to a third party. “That’s one of the things we wanted to prevent,” Thaine said. “The closer you are to the source of the data before you strip out the personal data the better it is.”
Private AI’s tool consists of three lines of code that operates directly within its clients’ workflows and infrastructure, preventing sensitive data from being shared outside of a client’s system.
“We believe that privacy is at the beginning of its moment, and finding founders who are visionary in that space and building superior technology got us excited,” explained Jonah Midanik, managing director at Forum.
“When we met Patricia, she clearly explained why her AI is better than anything else out there, and we immediately knew we wanted to invest,” Midanik said. “After spending 16 intense weeks together in our accelerator program, and watching their team develop for huge customers like Twilio as well as continue to attract global top talent while building their company in Canada, we were certain we wanted to continue to invest in their seed round.”