Washington DC - The National Highway Traffic Safety Administration said it was deepening its investigation into Tesla's "self-driving system" over concerns about its performance in poor visibility.

Analysis of crash data suggests the technology may have problems under such conditions, the NHTSA said on Thursday.

Tesla CEO Elon Musk has pushed for vehicles to operate autonomously using only cameras as sensors. Competitors and many experts have warned that cameras alone may not be sufficiently reliable, particularly in difficult visibility conditions.

Other developers of self-driving vehicles, including robotaxi leader Waymo, a subsidiary of Google parent Alphabet, also use more expensive technologies such as laser radar systems to scan a vehicle's surroundings.

The NHTSA is examining two versions of the system, known as Full Self-Driving (FSD). Despite the name, it currently functions only as a driver assistance system, with drivers required to remain in control at all times.

Tesla refers to one version as "supervised" FSD, which must be monitored by users. A more advanced version, in which vehicles are intended to operate autonomously, is so far available in the US only as a beta test.

According to the NHTSA, Tesla has said the FSD system is designed to detect when cameras encounter visibility issues and to alert drivers. However, in the crashes under review, the system did not recognize when cameras were impaired – for example by glare or airborne particles – and did not issue warnings.

As a result, drivers did not have sufficient time to react, the agency said. The investigation covers Tesla models produced from 2016 onwards.