#United States #Mountain View #AI Security #WitnessAI #NLP

WitnessAI Unveils NER-D: Transforming Sensitive Data Detection

WitnessAI, known for its innovative AI-native security platform, has introduced a groundbreaking model called NER-D, which stands for Named Entity Recognition - Double pass. This cutting-edge technology is designed to identify sensitive information in artificial intelligence conversations by emphasizing the meaning and context of the words used rather than merely relying on static patterns.

A Paradigm Shift in AI Speed and Accuracy

Setting apart NER-D from existing solutions, this model successfully eliminates the long-standing trade-off between speed and accuracy that has challenged the industry. With remarkable speed, NER-D can classify information in a single parallel pass, achieving a performance level that is twenty times faster than traditional generative methods without sacrificing accuracy. This presents a significant leap forward, particularly in real-time applications where timely data protection is crucial.

The capability of NER-D enables artificial intelligence to comprehend the context behind various terms. For instance, during an active conversation, it can quickly determine whether “Paris” refers to a city in France, a public personality like Paris Hilton, or a confidential project codename. This intelligent contextual understanding is critical in a landscape where legacy systems fail to deliver effective nuanced data protection.

Enhancing Contextual Detection in AI Conversations

The NER-D model signifies a shift toward full-context semantic data protection at real-time speeds. It evaluates conversations similarly to human reviewers, identifying sensitive information based on contextual meanings instead of merely their structural forms. By integrating large language model (LLM) knowledge with operational speed, NER-D fills the existing gaps left by older solutions. It simplifies the securing of sensitive data, enabling enterprises to protect information just by describing the concept of data involved, without the need to manually catalog variations.

Rick Caccia, the CEO of WitnessAI, emphasized the core mission of their organization: “Our goal is to secure enterprise AI. Traditional tools leave substantial blind spots with unstructured and proprietary data, and NER-D addresses this critical need for robust protection.”

Key Innovations of NER-D

The distinctive features of the NER-D model include:

• - Benchmark-leading Accuracy: With a performance level that exceeds prior achievements by 7.9 points, NER-D achieves the highest accuracy among all current detection methods. Its strength lies in its handling of complex data types, where contextual awareness is pivotal.
• - Production-Ready Speed: Rapid classification is made possible through one parallel pass, allowing NER-D to operate over twenty times quicker than comparable generative methods.
• - Robust Knowledge Base: Grounded in comprehensive understanding, NER-D is adept at recognizing global data formats and industry-specific concepts, such as national ID formats and pharmaceutical indicators. This facilitates the effective security of sensitive information merely by concept description.
• - Reduction in False Positives: NER-D assesses the context of conversations to infer true intent, thus significantly lowering the rate of false alarms commonly experienced by traditional tools. For instance, it can discern benign numerical entries, like purchase order values, from sensitive identification metrics.

According to Amr Ali, Head of Machine Learning at WitnessAI, “While legacy tools merely match patterns, NER-D assesses the genuine meaning of conversations in real-time. The race for AI data protection hinges on understanding context and recognizing meaning, and NER-D marks a significant advancement in this pursuit.”

The NER-D feature will be integrated into the WitnessAI platform in the upcoming months, seamlessly fitting into existing data protection workflows, such as redaction and tokenization.

For further details regarding WitnessAI or to schedule a demo of NER-D, interested parties are encouraged to visit WitnessAI's website.

About WitnessAI

WitnessAI is the trusted AI security platform for enterprises aiming to govern and safeguard AI-related activities. It provides unparalleled visibility into all interactions, including those involving autonomous agents, in native applications where older tools have limitations. Unlike conventional methods that depend on outdated keyword recognition, WitnessAI’s AI-centric platform aims to understand intent, context, and underlying meaning, enabling intelligent protocols to mitigate modern threats such as prompt injection while bolstering productivity. Its enterprise-first, single-tenant framework guarantees data sovereignty and compliance, transforming security protocols from obstacles into facilitators of AI strategy.