Sentra has launched an AI classification capability designed to identify sensitive information within unstructured data at enterprise scale. The vendor claims its specialised Small Language Models achieve 99% accuracy while processing petabytes of data.

A new AI Classifier for Unstructured Data capability uses domain-specific SLMs rather than general-purpose Large Language Models. According to the vendor, this reduces computational requirements by an order of magnitude compared to LLM-based classification systems deployed by competitors.

“Generative AI offers incredible potential, but it also introduces new data risks if sensitive or proprietary information isn’t properly governed,” said Yair Cohen, VP, Product and Co-Founder of Sentra.

“Our AI classification engine provides unmatched scale, accuracy, context and adaptability - empowering enterprises to innovate with GenAI safely, responsibly and cost-effectively.”

Each SLM-based classifier is ‘domain optimized’ to surface the full context of unstructured assets and files by categorizing departments (such as finance, legal, HR and marketing, etc.), industry (healthcare, retail, etc.), geography, ownership (employee, customer, etc.), sensitivity and company-specific intellectual property.

Sentra claims this provides context beyond pattern-matching approaches used by traditional DSPM solutions.

After initial scanning, the platform automatically learns to identify proprietary data categories without requiring rescans. The system supports more than 70 languages and can identify specialised categories including pharmaceutical research data and proprietary formulas.

The announcement positions Sentra against established DSPM vendors including Cyera, BigID, Varonis and others in a market projected to reach US$415 million in 2024 revenue.

https://www.sentra.io