Chapter 2 · EV09

TfL x Uber: Reporting Failures and the Limits of Operator Oversight

Date collected: 26 May 2026

Summary

TfL revoked Uber London licence over unreported serious incidents; operator self-reporting gaps expose structural limits of regulator-dependent oversight

The 2019 licence revocation

TfL declined to renew Uber’s private hire operator licence in November 2019, finding Uber “not fit and proper”. The specific triggers:

A change to Uber’s systems allowed unauthorised drivers to upload their photos to other Uber driver accounts — enabling them to collect passengers while posing as the booked driver. This occurred in at least 14,000 trips.
Dismissed or suspended drivers were able to create new Uber accounts and continue carrying passengers.

The reporting failure (the institutional ODD point)

TfL’s scrutiny identified that Uber did not report these breaches as safety matters, and did not escalate them to TfL’s Taxi and Private Hire team through the required written safety notification channel. Uber’s own biannual Assurance Reports — the mechanism by which TfL was supposed to exercise ongoing oversight — failed to recognise the importance of the breaches that had occurred.

TfL found the Assurance Reports “less useful than they might have been.” The operator’s reporting framework did not align with the regulator’s oversight requirements.

What makes this relevant to CAM

This is a direct precedent for the data-and-reporting domain of Institutional ODD. The problem was not that Uber’s vehicles were failing technically. The problem was that the operator’s reporting systems were misaligned with what the regulator needed in order to maintain effective oversight. The city’s ability to govern the system depended entirely on data and disclosures from the operator — and that data was insufficient.

Note for Box 2.3 (London): The TfL/Uber case is a stronger evidence point for the reporting/governance domain than for “data access disputes” in the sense of raw operational data. The framing should be: London’s regulatory experience has already demonstrated that the city’s ability to govern mobility operators depends on the quality and completeness of operator reporting — a lesson directly applicable to CAM oversight.

Cite as: TfL Decision Letter to ULL 25 November 2019; CNBC; TechCrunch

Source links:

Drafting Notes for Ch.2

For 2.2 (What is being deployed) — replace abstract framing with:

As of early 2026, the most commercially advanced deployment is Waymo’s US operation: approximately 2,500 vehicles across five cities, completing over 14 million trips in 2025 alone — a 3× increase on 2024 — and exceeding 450,000 weekly rides by December 2025. That growth trajectory (from 10,000 weekly rides in May 2023 to 450,000 in 24 months) illustrates the speed at which commercially viable scale can arrive. In China, more than 20 cities have authorised Level 4 trials. Europe has seen far fewer commercial deployments, but has developed more governance scaffolding in advance.

For Box 2.2 (San Francisco) — add the specific governance failure detail: The key fact is not the pedestrian incident but the reporting gap: Cruise withheld footage and did not classify the vehicle’s subsequent movement as a safety event requiring regulatory notification. The city’s oversight capacity depended entirely on the operator’s own disclosure — and that disclosure was incomplete. This is the institutional failure, not the technical one.

For Box 2.3 (London) — add TfL/Uber reporting failure precedent: London’s most instructive recent case is not a CAM case but the TfL/Uber licence revocations (2019, 2023). In both instances, TfL’s ability to exercise effective oversight was undermined by the operator’s reporting frameworks — not by an absence of regulatory will. 14,000 trips with unauthorised drivers proceeded undetected because the operator’s systems did not surface the problem and the reporting channel did not require notification. This is a direct governance-architecture lesson for CAM deployment.

Thread