How Ascentt Implemented AI to Protect PII in Vehicle Testing

Artificial Intelligence is everywhere, but generic, horizontal AI models do not always work, especially in industries like automotive, where use cases are unique. Specialized, domain-specific AI systems are key to solving complex, high-value problems in scenarios like connected vehicle testing. Vertical AI is now a blueprint for competitive advantage in the automotive industry.

More notably, it can capture (and mask) personally identifiable information (PII), such as faces, license plates, house numbers, and any data that can be tied back to an individual. However, this introduces a compliance challenge: How can vertical AI models maintain compliance with strict privacy laws like GDPR, CCPA, or PIPL? The answer lies in rethinking privacy as a design pillar, not a blocker.

Read further to uncover how Ascentt built a real-time PII obfuscation AI agent using deep learning and signal processing, and why this solution reflects our broader commitment to Vertical AI in the automotive industry.

The PII Compliance Challenge

Test vehicles collect millions of terabytes of data every year. But unlike staged environments, real-world driving footage is unpredictable. Pedestrians walking by. License plates of nearby cars. Storefronts and street addresses. All of it can potentially violate regional privacy laws if not handled correctly.

And while many traditional obfuscation methods exist, they’re:

• Slow (introducing lag in global data sharing).
• Error-prone (missing faces or plates).
• Unscalable (manual QA and region-specific rules make it operationally complex).

This presents a significant bottleneck for automakers and suppliers trying to innovate across borders.

Why Horizontal AI Falls Short

Out-of-the-box horizontal models may detect a face, but might not be able to distinguish between a bystander and a billboard photo. They might also not accurately adapt obfuscation to preserve key automotive elements like sensor alignment or bounding boxes.

In the automotive sector, teams aren’t just working with “video”. They’re working with multi-modal, synchronized, time-sensitive streams where even a blurred pixel can interfere with sensor calibration or ML model performance.

Horizontal AI models struggle to:

• Process and analyze high-frequency connected data
• Navigate the complex product configuration lifecycle
• Adapt to shifting user behavior with EVs and autonomy

The Bigger Picture: Vertical AI in Automotive

Vertical AI models are trained on automotive datasets and tuned for sensor and camera geometry. Integrating edge processing pipelines and automotive-grade computing units
redefines the future of automotive. These models have built-in domain intelligence, streamlined data flows, and are precision-tuned for industry-specific tasks.

At Ascentt, we’ve helped leading OEMs and Tier-1s solve problems such as:

• Predictive vehicle diagnostics using sensor fusion and temporal modeling.
• Process vision in manufacturing to reduce defects per unit.
• Data optimization for connected vehicle fleets, improving bandwidth and data
  quality tradeoffs.

Each solution is bespoke and engineered to fit within the exact data realities, business priorities, and client compliance frameworks. They all have something in common: they go beyond AI as a tech layer and embed intelligence deep into the domain.

Our Approach: Vertical AI for Automotive Compliance

At Ascentt, we specialize in Vertical AI, domain-specific, production-ready intelligence systems built for complex industries like automotive. Our approach goes beyond
adapting general-purpose AI tools; we build intelligence into the industry’s workflow, data formats, and operational logic.

So, how did we apply that philosophy to the privacy problem? Here’s a real-world case study.

The Problem

Due to privacy compliance risks, global test vehicle programs were stalled for a leading
automotive manufacturer. Teams needed a way to:

• Automatically obfuscate PII from video footage in real-time.
• Maintain complete engineering insight for computer vision tasks.
• Operate across regulatory environments without manual intervention.

The Solution

We designed and deployed a real-time video obfuscation agent, a vertically integrated AI module that sits inside the data pipeline of test vehicles or edge servers.

Key components included:

• Deep Learning for Detection: Specialized neural networks trained on diverse driving environments to detect faces, license plates, and other PII across lighting/weather conditions.
• Signal Processing for Obfuscation: Region-adaptive blurring and pixelation methods that balance privacy with visual utility, preserving lane markings, pedestrian outlines, and vehicle behavior cues.
• Rule-Based Privacy Layer: Built-in geographic toggles to comply with different jurisdictional rules (e.g., blurring entire pedestrians in the EU, license plates in the U.S., etc.)

The Result

The deep learning + signal processing pipeline we built resulted in:

• Removing PII in real-time, with near-zero latency.
• Unblocked data pipelines, enabling engineering teams across geographies to work from the same video assets.
• Stronger regulatory compliance without sacrificing model training quality or human QA effectiveness.

Turning Privacy into a Competitive Advantage

Privacy and compliance are often treated as checkboxes, something to deal with after the fact. But they can be a core differentiator if you design around them from the start. By integrating real-time PII obfuscation directly into the test vehicle pipeline, our clients ensured compliance, sped up global R&D, avoided legal risk, and enhanced public trust. This is the power of industry-aware AI; it turns red tape into green lights.

At Ascentt, we don’t build AI for the sake of AI. We build Vertical AI, and the real-time PII Obfuscation Agent is one of many ways we’re helping automotive leaders bridge the gap between innovation and compliance. In this new era, relevance beats generality every time.

Interested in deploying privacy-aware AI in your vehicle programs? Hear our Executive Vice President, Shailesh Shedge, discuss vertical AI’s role in the automotive and manufacturing industry.

Let’s discuss how Ascentt can help you create specialized data pipelines, build scalable, domain-informed AI agents, and turn PII regulation into a strategic asset.