Daily Tech Wizard Newsletter

Stories / / By Daily Tech Wizard

Wayve pitches AI driver whose safety can be measured

Wayve says no rulebook can cover every road, so its AI driver must learn

Wayve says an AI driver doesn't have to be a black box.

In a 60-second video posted November 6, the company pitched an AI driver whose behavior can be observed, measured, and validated. It calls the approach Learned Safety.

According to the video, the AI driver includes risk assessment in every decision. A component Wayve calls the learned safety envelope makes those safety judgments observable.

According to the video, two tools turn that behavior into evidence. Gaia provides closed-loop simulation. Grader provides learned evaluation. Wayve says together they let it measure and validate behavior at scale.

The reasoning is that no rule book can cover every scenario, so an AI driver has to generalize beyond what it has seen. Wayve has been developing end-to-end embodied AI since 2017 for the long tail of driving: situations that can't be anticipated through maps or hand-written rules.

As it brings its AI Driver to production with partners, Wayve says the question is no longer whether end-to-end AI can work, but how to build and validate safety at global scale.

One caveat: the video doesn't describe in detail how Gaia and Grader work, and it doesn't include results. The claims are Wayve's own.

Would you trust an AI driver more if you could see how it judges risk?

Video: Wayve on YouTube