A grainy, smiling male face appeared on hundreds of public digital signs and on-board screens across central Seoul at 08:17 on Tuesday, triggering facial‑recognition software used by a private bus operator and bringing 73 buses to an automatic stop. The stoppage snarled the morning commute in four districts, prompted a trading halt for ad‑tech firm Orbis Find and prompted an urgent probe by the National Personal Data Board (NPDB).
How the glitch played out
Bus drivers and passengers described a surreal sequence: on digital ad panels that double as ticketing and security interfaces, the unknown face was flagged as a high‑risk identity by Orbis Find’s FaceCast API. The bus systems are contracted to a private operator, BlueRoute Mobility, which uses FaceCast to speed boarding by matching subscribed passengers to permits. At multiple stops the matched flag caused doors to lock and the driver interface to display a 'security hold,' forcing crews to perform manual overrides.
"I pressed the panel and the doors wouldn't budge; the screen showed a man I didn't know and then all the lights went red," said Ada Abacı, a BlueRoute driver who was at the wheel of one halted vehicle near Gwanghwamun. "Commuters were calm but confused — no one knew whether to get off."
Reaction from regulators and firm
NPDB chair Park Soon‑hyun told reporters the board had opened an inquiry and was treating the event as a potential breach of personal‑data handling rules. "This was an unacceptable failure of a system that holds sensitive biometric markers," Park said. "We are considering administrative measures under the Personal Information Protection Act up to ₩620 billion."
Orbis Find issued a brief statement saying it had suspended FaceCast across public transport while engineers worked with BlueRoute and emergency services. Company CEO Han Jae‑min (an invented corporate name) said in a call with investors that the firm had isolated a malformed feature vector in an update pushed overnight, and that the firm had not intentionally used any individual’s image without consent. He declined to say whether the face belonged to any public figure.
It felt like the whole city froze because a face nobody recognised told the machines what to do.
Ada Abacı, bus driver
Market fall‑out and advertiser pullback
Trading resumed for Orbis Find but shares closed down 24%, wiping roughly ₩8.1 trillion (about €5.3bn) from its market value. Several major advertisers, including two retail chains and a global automaker, ordered immediate suspension of campaigns run through FaceCast; early estimates put short‑term ad revenue losses around ₩140 billion (≈€9.3m). "Advertisers are terrified of automated systems misclassifying customers and exposing them to reputational risk," said Kevon E. Mueller, a Seoul‑based media analyst. "This will take months to restore trust, if not longer."
Security engineers working with the NPDB said forensic traces of the file — a compressed facial feature payload — led to a cluster of archived video snippets uploaded last week to a small content farm. That link, they said, does not explain why the FaceCast matcher labelled the payload as a high‑risk 'match' that triggered lockdowns. BlueRoute said all halted buses were back in service by noon and no passengers were injured.
What happens next
The transport ministry has summoned Orbis Find and BlueRoute for questioning and scheduled an emergency parliamentary session for Friday. Consumer‑rights lawyers said they were preparing class actions on behalf of commuters. Orbis Find said it would roll back the overnight update, offer a temporary credit to public transport operators and cooperate with NPDB, but did not rule out further regulatory penalties.
Investigators still cannot account for who created the face payload. A tracing effort led to an old actor's publicity reel from the 1970s that circulated on a digital archive, but forensic analysts say the payload carried data transformations that do not match simple uploads or edits. Maia Yrjölä‑Helppolainen, a commuter whose phone recorded the screen, said: "We need rules so machines cannot decide for us — people should not wake up to find themselves labelled by a stranger's face."
