Recording your voice may seem harmless, but it may actually reveal your identity, as well as additional data about you, such as how you feel. But it can also detect diseases you may be suffering from.
People may not understand this yet, but companies that process data increasingly realize that they need to treat voice as personally identifiable information. This is especially true in Europe in the context of the GDPR: while many companies hope to build AI on top of voice data, in many cases this requires removing biometric information first.
This is where Nijta hopes to help: by providing AI-powered speech anonymization technology to customers who need to comply with privacy requirements. Although its name is a Hindi word meaning privacy, it is headquartered in Lille, France, where its Indian CEO, Brij Srivastava, has moved to pursue a PhD at INRIA, the French institute for research in computer science and automation.
Nijta was born into the Inria Startup Studio, a program aimed at supporting entrepreneurs with a PhD who want to start a business. It worked: Nijta is now an award-winning young B2B company with €2 million in funding from various sources, including French deep-tech venture capital fund Elaia and Lille-based investment firm Finovam Gestion.
“Europe is our core market,” Srivastava told TechCrunch. The main reason is simple: “The General Data Protection Regulation (GDPR) is a very strong data privacy law.” While voice anonymity can be relevant to many sectors, Nijta's Achilles heel is its combination of compliance and business opportunities.
“Nijta’s AI-powered voice anonymization technology provides a solution for many companies that are increasingly concerned about data privacy and excited about generative AI,” Celine Passeduit, investment director at Elaia, said in a statement.
Increasing use cases
Contact centers in general are potential customers for Nijta, but even more so when they deal with health data.
One of her early collaborations was around Oky Doky, a project aimed at better handling emergency medical calls. While it's easy to see how AI could help, it's clear that voices must be anonymized to remove speaker identity and personally identifiable information from the training data.
Other use cases include defense scenarios, which Srivastava did not expand on for obvious reasons, but also education technology, where children's voices must be anonymized before AI can be leveraged to give them feedback on pronunciation, for example.
Content created by Nijta is watermarked, which has become the standard if not the norm for all things generative AI. The startup also says that Nijta Voice Harbor's protections are irreversible, unlike some of the voice edits that media outlets unwisely use in hopes of protecting the victims they're interviewing.
Lack of awareness of voice-related privacy issues is one of the challenges that Nijta will have to face. This is also why starting with B2B and Europe makes sense: even if customers don't push for voice privacy, risking a hefty fine turns companies into early adopters.
Eventually, however, Nijta hopes to expand into business-to-consumer interactions, focusing on securing recorded messages, for example. “Real-time anonymization for secure communications is also something we are exploring very actively,” Srivastava said. But B2C is still a few years away; Niigta's small team can only spread itself too thin.
Northerly background wind
Negta has seven team members, including Srivastava, full-time co-founders Seyed Ahmed Hosseini and Nathalie Fouquier, and his former professor, senior research scientist and part-time co-founder Emmanuel Vincent. Srivastava hopes to grow the team to 10 people by June, but is also receiving outside help for efforts the startup wouldn't undertake on its own.
Business France, in particular, helps Negta reduce internationalization costs, Srivastava said. “Because we are a small company, we cannot employ many sales representatives in different countries.” Instead, it can rely on prospecting from a Business France representative in a particular country, “and the cost is mostly subsidized by [Lille’s] Hauts-de-France region.” Additionally, it opened doors for the startup in our sister state of Maryland.
This is one of the reasons why Srivastava has no trouble answering when he is asked (often) why Negta is in Lille, and not in Paris. While some of its tailwinds are more broadly linked to France, I have found that the country's far north is conveniently located in close proximity to Paris, but also Brussels, Amsterdam and London.
However, for Nijta to become global, it will have to become multilingual. This is a major R&D challenge, but one the startup is working on, with its eye on Europe and Asia. It is also helpful that the startup will receive another million euros in deep technology development aid from Bpifrance, a joint grant and repayable advance to finance R&D expenses; This will also make the question of why Srivastava chose Lille and France easier to answer.
