Weekly lists of undertrials who have served long enough to seek release
TL;DR Bailclock reads jail admission records and charge details, works out when each undertrial crosses the detention limit in the new criminal procedure code and hands legal aid lawyers a weekly list with the papers they need. District legal services authorities and prison departments pay. It predicts nothing about people.
At the end of 2022, India's prisons held 5,73,220 people at 131.4% of capacity, and 75.8% of them were undertrials who had not been convicted[1]. About 1,34,799 undertrials had been held for over a year[2].
The law already offers a way out for some of them. Yet in 2022 the official count found only 1,487 undertrials eligible for release on time served, and state figures disagree wildly. Uttar Pradesh reported 30,415 eligible to Parliament over 2017–2021 but 5,825 to the national crime bureau[2]. Nobody is reliably doing the arithmetic, prisoner by prisoner.
What it does, in regulatory terms: a record-keeping and alert tool. It gives no legal advice, and a lawyer checks every flag before anything is filed.
The seed proposed AI that predicts who will reoffend, and I dropped that without a second thought. ProPublica's study of a US risk tool found Black defendants who did not reoffend were labelled high risk 44.9% of the time, against 23.5% for white defendants[3]. India's recorded recidivism is 1.9% of convicts admitted[1], and the real overload is people waiting for trial.
Bailclock takes each undertrial's admission date and charges, looks up the maximum sentence and applies section 479 of the Bharatiya Nagarik Suraksha Sanhita - release after half the maximum, or one-third for first-time offenders, with its exclusions[4]. It flags people approaching or past the limit, those granted bail but still held, and records with missing data. The panel lawyer gets a ranked list and a draft application pack.
The users are legal aid lawyers, jail legal-aid clinic staff and jail superintendents, who now carry the duty to identify eligible prisoners[2]. The payers would be state legal services authorities or prison departments, through a state contract. Prison-reform NGOs and philanthropic funders are a realistic first customer, since government budgets are thin. There's no public price for a comparable tool.
The new criminal procedure code took effect in July 2024 and added the one-third rule for first-time offenders, which widens who can apply[4]. Prison records are now digital too, and the national ePrisons portal covers 1,359 prisons and about 23.5 million records[5]. So the data exists but the eligibility calculation doesn't.
There's no public market figure. The demand signal is the gap - between 3,800 and 5,100 prisoners a quarter in 2024–2025 had been granted bail but stayed in jail because they could not furnish surety[2].
Correctional staff such as welfare officers were also 44% short of sanctioned strength in 2022[1], so nobody has spare hands for this.
ePrisons runs prison records and a legal aid login, but I found no public eligibility alert[5]. Under Trial Review Committees in each district review cases each quarter, working from the same patchy counts[2]. Prison-reform NGOs do this by hand in some jails.
The gap is software that does the calculation every week for every prisoner and shows its working.
| Difficulty | Moderate (46/100) |
| Starting capital | ₹35–70 L |
| Time to launch | 6–12 months |
The software is simple. The hard parts are getting access to jail records, mapping thousands of offences under old and new laws to maximum sentences and selling to state bodies that rarely buy software of any kind. Plan on grant money for the first year or two.
No licence covers the software itself, but legal and data rules decide what it can do.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Bailclock is rules-based on purpose, so AI plays a supporting role.
Never let AI decide who qualifies or predict anyone's behaviour. The eligibility calculation stays as fixed rules that show their working, and a lawyer checks every flag.
Want to work out the AI layer for Bailclock? Talk to Yash →
Here's what can go south -
I think this is a strong idea for a founder with legal-tech or justice-system experience who can work with government and NGO funders. If you need venture-scale returns then it's a poor fit.
Prove one thing first - in one district jail, show the tool finds eligible prisoners the current process missed.
My estimate is ₹35–70 L to launch, mostly for offence mapping, secure hosting and the long wait for state access.
No licence, but you need written permission to read jail records, a processor contract under the DPDP Act, and output that goes only to enrolled lawyers.
Possibly, wherever pre-trial detention limits exist and records are digital. The rules and data access process differ a lot, so start with the local law.
The free Bailclock validation kit has the line-by-line capital breakdown, the effort scoring, the data and legal checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for legal aid lawyers and jail staff, and 5 assumptions to test. Download the kit →
Thinking about building Bailclock or something close to it? The make-or-break work is winning one state's permission to read jail records, and I'm glad to help you work out how to ask. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.