Inside Uber before its self-driving car killed a pedestrian: sources describe infighting, “perverse” incentives, and questionable decisions
Julie Bort / Business Insider : Tweets: @nichcarlson Tweets: Nicholas Carlson / @nichcarlson : “Sources say engineers were pressured to “tune” the self-driving car for a smoother ride in preparation of a big, planned year-end demonstration of their progress. But that meant not allowing the car to respond to everything it saw, real or not.” http://www.businessinsider.com/ ...
Context & Ripple Effects
This piece extends a paper trail that began with Recode's 2017 reporting on Uber's Advanced Technologies Group — slow technical progress, haphazard public demos, and an ATG talent exodus triggered by the Otto acquisition. The NTSB's May finding that the vehicle failed to identify the pedestrian or brake until it was too late established what went wrong mechanically; Business Insider now supplies the organizational why.
The new sourcing alleges engineers were pressured to tune the car for a smoother ride ahead of a planned year-end demonstration, suppressing responses to hazards it detected — meaning the fatal March crash happened inside a program already trading safety margin for showmanship.
First-order effects
- Uber's ATG engineers faced incentives that rewarded a polished demo over hazard response, directly degrading the vehicle's safety behavior in the period before the pedestrian was killed.
- The reporting hands regulators and litigants a documented motive structure — 'perverse' incentives plus infighting — that maps onto the NTSB's mechanical findings of failure to identify and brake.
Second-order effects
- When Uber moved to resume autonomous-mode testing in Pittsburgh weeks later, sources told the New York Times the cars still had unresolved issues, so the internal problems described here carried straight into the restart decision.
- Internal dissent escalated rather than subsided: an operations manager sent executives an 890-word email flagging serious concerns about the program just days before the crash, showing warnings existed above the engineering level.
Third-order effects
- If the pattern holds across Uber's safety apparatus — the Washington Post later found its Special Investigations Unit often put Uber's interests ahead of riders' — the structural lesson is that self-regulated incident review cannot police a culture tuned for demonstrations, strengthening the case for external oversight of AV testing.
- For the industry, the episode makes demo-driven development cadence a liability model: any autonomy program whose milestones are staged public showcases inherits the same incentive to mute the system's caution.
The trend: Autonomous vehicle programs are colliding with the gap between demonstration-driven incentives and safety governance, pushing incident review from internal units toward regulators.