AI That Predicts Car Crashes Before They Happen

Imagine if your car’s AI could warn you about a pedestrian stepping into the street *before* you even see them. That’s the idea behind PRISM, a new AI safety system for self-driving cars. Most current car safety systems only react after a hazard appears—like slamming the brakes when a cyclist suddenly swerves. PRISM flips that script by predicting danger before it happens. It works by combining three AI models: one tracks how cars and bikes move, another watches the environment (like slippery roads), and the third focuses on pedestrians and cyclists.

PRISM doesn’t just guess—it scores risks like a weather app rates storm chances. In tests with 1,296 real-world driving scenarios, it gave an average safety score of 68 out of 100 and flagged 77.6% of risky situations as advisory. It even caught 3.8% as near-misses and escalated 11% to alerts or emergency responses. The system learns on the fly, adjusting like a thermostat for safety. It’s like having a co-pilot who’s always one step ahead.

Why does this matter? Because self-driving cars are coming, and safety is the biggest worry. PRISM could reduce crashes with pedestrians and cyclists, who are often the most vulnerable. It’s also designed for busy cities, where unpredictability makes driving trickier. The researchers say it’s a step toward cars that don’t just avoid crashes—they prevent them.

Right now, this is still research, not a product. But it shows how AI could make roads safer for everyone, not just drivers.