Posted on the 3rd July 2026 in PHTM News

LONDON ROBOTAXI PUSH SPARKS ALARM OVER HIDDEN AI BIASES AND PASSENGER SAFETY DEFICIENCIES

Plans to launch driverless taxi services across London later this year have come under intense scrutiny after artificial intelligence experts warned the vehicles might fail to spot pedestrians depending on what they are wearing. The revelations emerged during a recent London Assembly meeting, raising serious doubts about whether the automated technology is truly ready for the capital's complex streets.

Autonomous vehicle (AV) operators, including tech giant Waymo, are currently conducting human-supervised testing with the goal of rolling out commercial passenger services in London as early as September 2026. However, safety specialists have flagged alarming blind spots in how these vehicles recognise people.

Professor Siddartha Khastgir, Head of Safe Autonomy at the University of Warwick, revealed that experimental evidence shows sensors can easily fail based on a person's appearance. "We have experimental evidence from our collaborators in Canada who have shown to us that, depending on the clothing of the pedestrian, the sensors may or may not detect the pedestrian," Khastgir said, noting that thick winter clothing is a known trigger.

Further academic data underscores systemic biases within the software. Research from King’s College London reveals that AVs are roughly 20% more likely to detect adults than children, and just over 7.5% more likely to spot white people than ethnic minorities.

This discrepancy stems from flawed training data, as the open-source image galleries used to programme the vehicles do not accurately represent a diverse public. Prof Khastgir stressed that training data must include variables ranging from skin tone and hair colour to reflective clothing and summer shorts so that "this bias is not part of the detection process."

Beyond external road hazards, the assembly meeting exposed sharp disagreements regarding in-cabin passenger security.

When asked how a driverless car would protect a passenger from assault or harassment, Ben Loewenstein, Waymo's Head of Policy and Government Affairs for the UK and Europe, suggested that cabin movement or an undone seatbelt would alert a remote team member to "beam into the car" and check on passengers or call emergency services.

Labour's Transport Spokesperson on the London Assembly, Elly Baker AM, argued that a remote monitor cannot replace the intuition of a human driver who can spot uncomfortable, quiet interactions that do not involve major thrashing or seatbelts being undone, adding: “I’m not hearing the level of concern about how serious this is at this stage.”

Baker also voiced strong concerns about the lack of transparency regarding the technology's broader impact on public transport and the livelihoods of professional human drivers. “What we heard raises serious doubts about whether this technology is actually ready for London," she stated.

Waymo has since pushed back against the criticism, with a spokesperson strongly denying any suggestion that the company is dismissing the threat of abuse.

"Safeguarding is critically important, and strong protections and support for passengers are at the core of our operations," the spokesperson insisted.

The company maintained that any reports of threatening behavior are met with robust protocols, including 24/7 human rider support and direct emergency service access, promising to continue working with Transport for London and safeguarding experts to ensure strict protections are established from day one.


Read another story