Kodiak AI says it remains on track to launch driverless long-haul commercial trucking operations by the end of 2026.

The autonomous-truck developer said its internal measure of readiness for driverless highway operations reached 93% at the end of August, up from 84% in February.

Kodiak calls the metric its Autonomy Readiness Measure, or ARM. It represents the percentage of claims and supporting evidence in the company’s long-haul safety case that Kodiak considers materially complete.

A safety case is the body of evidence and analysis intended to demonstrate that an autonomous driving system can operate safely within a defined operating environment.

Kodiak said the remaining work is focused primarily on final engineering verification and validation.

“What makes this measure powerful is that we know exactly what stands between us and driver-out long-haul operations, and that list continues to get shorter month after month,” said Kodiak founder and CEO Don Burnette.

“We have identified and scheduled the remaining work,” he said. “We expect to complete the safety case before year’s end and launch driverless long-haul service.”

Building on Driverless Permian Operations

Kodiak already is operating driverless commercial trucks in the Permian Basin, where the company said 35 trucks were running with no humans in the cab at the end of the second quarter.

Those operations use the same core Kodiak Driver autonomous-driving architecture planned for long-haul trucking, although Kodiak develops a separate safety case for each operating environment.

The company’s highway safety case builds on lessons from its Permian Basic operations, but it has to address the different conditions and risks involved in long-haul highway freight.

Kodiak has reported steady progress on its long-haul ARM throughout 2026, moving from 84% in February to 86% in April, 91% in July and 93% in August.

Frank Fratrik, Kodiak’s director of safety, said the company’s confidence in removing the driver comes from hazard analysis, engineering safeguards and continued validation of the assumptions underlying its safety case.

“The confidence to remove the driver is something a program builds, and we have built ours deliberately,” Fratrik said.