There is barely a family in India that does not carry a crash story. A father who never came home. A cousin whose leg never healed right. A child hit outside a school gate. A delivery rider who left the news cycle long before his family recovered.

Mine is from 2005. I was at IIT Bombay. One night a batchmate and I were riding a Kinetic back late through Mumbai. He was driving, I was on the pillion. We did not see a divider in time. The scooter hit it and he was thrown across the road. I fell on the near side. A taxi ran over him. We picked him up and I held him in my lap the whole way to the hospital. He did not survive. I walked away without a scratch.

What stayed with me was not only the grief. It was the unfairness. Same vehicle, same night, same crash. One of us lived and one did not.

That is the strange place road death occupies in Indian life. Intimate enough to wound almost every family, abstract enough to stay normal, and fragmented enough that the system rarely learns.

In 2023 the Ministry of Road Transport and Highways recorded 172,890 road deaths, the highest India has ever logged. Even that is almost certainly an undercount. The WHO once estimated India's true toll at roughly 300,000 in 2016, about double the official figure for that year, and the government's own Sample Registration System put 2022 deaths far above the police count. The exact number matters. What matters more is that we still cannot agree on it. A country that cannot count its road deaths cleanly is not just struggling with traffic. It is struggling with truth.

The conversation stays at the visible layer: overspeeding, drunk driving, bad roads, missing helmets, weak enforcement. All of it is real. None of it reaches the deeper failure, which is trust.

I do not mean trust as optimism or civic niceness. I mean it operationally. Can institutions trust the data they are using. Can a bystander trust that helping a victim will not become a punishment. Can a family trust that compensation is real and reachable. Can the engineer, the police officer, the hospital, and the judge trust they are looking at the same picture of what happened. Too often the answer is no. And a system that cannot trust its own picture of a crash cannot learn anything from it.

This is not a fringe reading. Piyush Tewari founded the SaveLIFE Foundation after losing a teenage cousin to a crash, and rebuilt a stretch of the Mumbai-Pune Expressway on what engineers call the Safe System approach: the premise that people are fallible, will make fatal mistakes no matter how well trained, and that the system around them has to absorb the error. By fixing more than 3,500 engineering defects and cutting ambulance response from about 35 minutes to under 10, that single corridor saw deaths fall by roughly half between 2016 and 2020. The lesson is not that drivers there suddenly got better. It is that the system stopped letting their mistakes turn fatal.

I did not arrive at this from a safe distance. Neither did Amar Srivastava and Deepanshu Gupta, who founded the Indian Road Safety Campaign and built Crashfree India with Gajendra Jangid, my co-founder at Cars24. The last eighteen months of that work have shown the failure at close range, and the same pattern keeps surfacing: the breakdowns are connected.

The first failure is memory

Start with the data. India runs several road-death systems that do not reconcile: police-based MoRTH and NCRB figures, the survey-based Sample Registration System, the WHO's modeled estimates. The same crash should feed the same system, and the picture still diverges, sometimes by more than a lakh of deaths. A wrong baseline makes every target performative and every evaluation a guess. Digitizing the paperwork does not fix this on its own. Systems like iRAD and eDAR point the right way, a single geo-tagged record beats disconnected forms. But bad diagnosis can be digitized too. If an official is trained to close a file rather than investigate a chain of failure, faster data entry only speeds up bad judgment.

Roads everyone knows are dangerous

Then there is the road itself. Crashfree's Project Rakshak, a youth-led infrastructure effort, works on the part India routinely ignores: the places locals already know are unsafe, which formal systems act on only after enough people have been hurt or killed. A blackspot should not have to earn its name through repeated death. By mid-2026, Rakshak teams had assessed more than 120 high-risk locations across 18 cities and secured approvals for fixes at 25 of them, sites that had likely already seen more than a thousand crashes once under-reporting is factored in, with the interventions expected to cut crash risk there by a fifth to a half. The number matters less as scale than as method: dangerous roads can be made legible to authorities before more death forces their attention. The work also found that crash data on these sites is almost certainly undercounted, because non-fatal crashes vanish from formal memory entirely. The road was telling one story. The file was telling another.

After the crash, opacity keeps killing

The same break appears after impact, and it is the most under-discussed failure in Indian road safety. A country can pass a Good Samaritan law (India's was established by the Supreme Court in 2016 and later written into Section 134A of the Motor Vehicles Act to shield bystanders who help), publish ambulance numbers, and announce compensation frameworks. But if citizens do not know their rights, hospitals behave unpredictably, and families have no way to navigate the system, the law on paper builds no trust in practice. You see it in the first minutes after a crash, when a crowd gathers and no one is sure who to call or what protecting the victim will cost them.

This is why even a first-responder card matters. Early Crashfree material distilled roadside action into three words: Check, Call, Care. Make the scene safe, call 112, control the bleeding, start CPR if needed. In a working system that would be obvious. In India that clarity is itself missing infrastructure. Helping a victim should not demand private courage and improvisation every single time.

You see it again in trauma care. A visit to the AIIMS Trauma Centre makes the abstraction concrete. By Crashfree's published account, AIIMS sees roughly 125 to 175 trauma patients a day, more than 70,000 a year, and many serious crash victims still arrive with no meaningful pre-hospital care. Ordinary people, often rickshaw drivers, end up acting as first responders with no training and no equipment. Even at a leading trauma centre, the pre-arrival chain is strikingly thin. This is why PM RAHAT, the government's cashless-treatment scheme for crash victims, matters. Its strongest form is not compensation after the fact but treatment certainty in the minutes after impact: hospitals ready, police and hospitals synchronized, zero out-of-pocket cost for eligible victims. A crash should not turn fatal because the chain is confused about who admits, who pays, and who moves first.

Then compensation, where the law exists and access is broken. The Supreme Court said as much in January 2025, observing that a negligible number of victims benefit from the schemes because public awareness is low and dissemination by authorities is weak. A 2021 World Bank study found that among poor households hit by a crash, only 14% received any compensation at all, and the financial blow came to more than seven months of household income. Crashfree's field brief, Justice Unserved, shows the machinery underneath. Hit-and-run claims run at roughly 12% of eligible crashes a year. Seventy percent of surveyed low-income households did not know compensation existed. Nine in ten pending hit-and-run claims sit stalled on paperwork, while hundreds of crores lie unclaimed in the tribunals, ₹459 crore in Bombay alone, against a national backlog of more than ten lakh cases worth over ₹80,000 crore.

And into that vacuum step the middlemen. Unregulated "ambulance chasers" intercept victims at hospitals and sell them help with claims for a commission; in 2021 the Bar Council suspended 28 advocates for filing fake motor-accident claims. Where a system is opaque, trust gets privatized and sold back to the victim. India does not mainly have a compensation-law problem. It has a compensation-access problem.

The fix is often smaller than a new law. Many families do not know a claim must be filed within six months. The biggest unlock turns out to be a 90-minute awareness session and a printed checklist; at helpdesks run inside AIIMS and Sharda hospital, more than a hundred victims' families queued for exactly that. That is what opacity looks like up close. People do not only need money. They need translation. Policy says a victim can claim. Trust means the victim knows it, believes it, can prove it, and can survive the process long enough to receive it.

Nobody owns the chain

The failure to learn is also cultural. India still treats road death as the natural tax of growth: more vehicles, wider roads, rising fatalities for a while, and a quiet shrug underneath it all. We have built speed, volume, and aspiration. We have not built a safety philosophy equal to them. A line kept recurring in the work: information is infrastructure. Weak crash data makes safety design weak, invisible hazards make prevention weak, an illegible compensation system makes justice weak. And the deeper reason all of this persists is fragmentation. Road safety in India is split across transport departments, police, hospitals, local governments, courts, insurers, and ministries. Each can point to its own slice and still leave the person on the road inside a failure that nobody owns end to end. That is why India produces so much sincere road-safety activity and so little visible outcome. Everyone owns a fragment. Too few own the chain.

A mobility company that wants the upside of the category cannot keep treating the downside as somebody else's ministry. That is why this cannot be a CSR sidecar or an awareness week. Any company that benefits from mobility at scale inherits some responsibility for making the system more legible, more responsive, or less lethal. Safety is not adjacent to the category. It is part of the product.

None of this means Crashfree has cracked the problem. It has not. But by mid-2026 it had at least shown the outline of a different model: find dangerous infrastructure before it accumulates more death, treat compensation as a justice system rather than a legal footnote, build public tools instead of only reports, and use research to move institutions rather than decorate them.

A mature road-safety system assumes human beings will make mistakes. Its job is to make those mistakes less likely, less severe, easier to respond to, and easier to learn from.

Safer roads do not begin when drivers become perfect. They begin when systems can see, respond, and learn fast enough that imperfection is no longer so often fatal.

Notes and Sources