PVML Plans To Bring A New Era Of Data Protection In Artificial Intelligence

As artificial intelligence (AI) and generative artificial intelligence (GenAI) continue gaining popularity among businesses of all sizes, collecting and hoarding data has become even more prominent.

This is because models trained on more data end up becoming more efficient and offer better chances of success to their users.

At the same time, businesses must ensure that their data is well protected especially for customer data, personally identifiable information (PII) and sensitive information of individuals, to ensure data breaches and leaks don't end up being costly. Yet, ensuring high amounts of data access is important to model training. 

To solve this conundrum, Tel Aviv-based startup PVML has come up with a solution that enhances the safety of sensitive information, while allowing data sharing and enhanced analysis through a chatbot.

So, how does the company plan to deliver on such beneficial functions? Let’s explore!
 

What Is PVML Doing?

 

• PVML recently announced it had closed an $8 million funding round during the second half of 2023, which was led by venture capital firm NFX and supported by Gefen Capital and FJ-Labs.

• PVML was founded in 2021 by husband-and-wife team Shachar Schnapp (doctorate in computer science, specializing in differential privacy and having worked with General Motors) and Rina Galperin (master's in computer science with a focus on AI and NLP (Natural Language Processing) and having worked with Microsoft).

• PVML aims to bring “a new era of data protection” with a “data access platform engineered for the age of AI,” according to its website.

• It plans to do so by bringing an AI tool that’s stylized like ChatGPT for data analysis, blended with differential privacy, a framework that enhances data safety and privacy by analyzing data without revealing sensitive information about individuals in a dataset.

• This enables businesses to use data without having to make changes to original data, which also reduces overheads.

• The idea is enhanced by PVML using retrieval-augmented generation (RAG) to access data without moving it. This also helps bring hallucinations down to near zero.

• The platform’s data analysis capabilities are further enhanced by allowing users to “chat” with their data to draw insights without the risk of sensitive data finding its way in the chat.

• While differential privacy isn’t a new concept and is used by all large tech companies in some form, PVML feels it hasn’t yet been put into use by most of the data community. Additionally, data access solutions that are being used create large overheads.

• Differential privacy can also allow businesses to share data between business units and monetize sharing to third parties, according to the company.

 

What Did Stakeholders Say?

 

• Rina Galperin, the CTO of PVML, said, “A lot of our experience in this domain came from our work in big corporates and large companies where we saw that things are not as efficient as we were hoping for as naïve students, perhaps.”

• [Contd.] “The main value that we want to bring organizations as PVML is democratizing data. This can only happen if you, on one hand, protect this very sensitive data, but, on the other hand, allow easy access to it, which today is synonymous with AI.

• [Contd.] “Everybody wants to analyze data using free text. It’s much easier, faster and more efficient — and our secret sauce, differential privacy, enables this integration very easily.”

• Shachar Schnapp, the CEO of PVML, said, “The current knowledge about differential privacy is more theoretical than practical. We decided to take it from theory to practice. And that’s exactly what we’ve done: We develop practical algorithms that work best on data in real-life scenarios.”

• Gigi Levy-Weiss, general partner and co-founder of NFX, said, “In the stock market today, 70% of transactions are made by AI. That’s a taste of things to come, and organizations who adopt AI today will be a step ahead tomorrow. But companies are afraid to connect their data to AI, because they fear the exposure — and for good reasons.”

• [Contd.] “PVML’s unique technology creates an invisible layer of protection and democratizes access to data, enabling monetization use cases today and paving the way for tomorrow.”

Do you PVML will be able to bring about a revolutionary change in the data storage, use and access processes of businesses, particularly to train and create AI tools?

Let us know in the comments below!