DiDi monitors 100% of its rides with AI: only 3% of 17,000 alerts required police, and the figure needs context

By Amexi · October 22, 2026.

DiDi says artificial intelligence models analyze 100% of the trips taken on its app. According to figures released by the company itself, of more than 3 billion recorded rides, only about 17,000 triggered specialized monitoring mechanisms: 0.00057% of the total.

Of those cases, 7% were assessed as potential incidents and just 3% required support from public security agencies. The rest, the company says, were handled by its own safety teams.

What's new is the integration with the authorities. Through Axon 911, an emergency management platform, an update called Live Support lets the C4, C5 and C7 command centers in 19 states, connected to the national 911 system, see the trip's location in real time along with the vehicle's details. It is triggered when the system detects an anomaly or when the rider or driver asks for help. DiDi also cites an agreement with the Secretariat of Security and Citizen Protection (SSPC) for the emergency button.

Other figures provided by the company: more than 40 safety features, including audio recording and encryption, daily facial verification for drivers and more than 2,000 dashcams in vehicles in Mexico City. Some 95% of trips use a four-digit PIN to confirm the passenger is getting into the right car. Reported investment in infrastructure and safety in 2025 exceeds 15 million dollars. In addition, the LERT portal handles requests from authorities around the clock; according to DiDi, about 80% of those requests have nothing to do with trips taken on the platform.

**Our take.** The 3% figure can be read two ways, and it is worth not settling for the more comfortable one. The favorable reading: a system that filters well, escalates only serious cases to the police and doesn't overwhelm them. The cautious reading: these are the company's own figures, with no external audit. We don't know how many incidents slipped past monitoring, how many alerts were false or how many were detected late. A low escalation rate is not the same as a low risk rate.

There is also a tension the report doesn't resolve. Monitoring 100% of trips with AI is, by definition, mass surveillance of riders and drivers. It is justified on safety grounds, and here the benefit is plausible, but it demands clear rules: what data is retained, for how long, who can access it and under what conditions. That last point weighs especially heavily given the LERT figure: if four out of five requests from authorities are unrelated to trips on the platform, DiDi is also serving as an information source for general investigations. That point deserves scrutiny, and the report doesn't detail the safeguards.

Overall, the industry is moving toward platforms that integrate with emergency services, and it is a reasonable evolution: a panic button that actually reaches someone who can help is worth more than one that only alerts the company. It is an example of AI with a concrete short-term benefit. The condition for it to stay that way is transparency: publishing verifiable metrics (false positives, response times, undetected incidents) and limits on data use. With that, preventive surveillance earns trust; without it, it merely asks for it.