The Bengaluru Traffic Police (BTP) introduced an ITMS in December 2022 to reduce manual checks and speed up fine collection. A recent case, where a rider was fined for a non‑existent pillion, exposed gaps in the system’s accuracy and sparked a debate on the limits of AI in policing.
Key Developments
- Mid‑September 2026: A Bengaluru motorist received an e‑challan for pillion riding without a helmet, though he travelled alone. The AI camera mistook a guitar strapped to his back for a passenger.
- Senior officers admitted the error stemmed from both AI mis‑classification and a manual review oversight.
- BTP claims the ITMS has nearly 99% accuracy, but internal data shows accuracy dropping to 80% for certain violations by the end of 2023.
- Contactless challans now constitute about 87% of all fines, with roughly 7.9 lakh issued monthly.
Important Facts
The BTP uses three channels for issuing contactless challans:
- ITMS: AI detects violations and sends images to the Traffic Management Centre (TMC).
- FTVR (FTVR): Officers manually photograph offenders.
- Public Eye: Citizens upload violation photos anonymously via the BTP app.
According to the ASTraM database, about 19,000 AI‑flagged challans are generated daily, with a daily challenge rate of only 0.13%. However, this low challenge rate does not equate to high AI accuracy.
Common AI errors include:
- Helmet‑related flags when riders briefly remove helmets at signals.
- Signal‑jumping flags for vehicles moving straight during a brief right‑turn pause.
- Faulty number‑plate detection due to font variations.
- Seat‑belt flags when a black shirt’s colour matches the belt.
Exam Relevance
Understanding AI‑based enforcement touches on several GS papers:
- GS2 – Polity & Governance: Role of technology in public administration, accountability of law‑enforcement agencies, and citizen‑state interaction.
- GS3 – Economy: Impact of automated fines on revenue, cost‑effectiveness of policing, and the economics of smart‑city initiatives.
- GS4 – Ethics: Balancing surveillance benefits with privacy concerns and ensuring fairness in AI decisions.
Way Forward
To improve the system, BTP officials suggest:
- Continuous retraining of AI models with diverse data (e.g., varied helmet colours, different number‑plate fonts).
- Stronger human oversight at the TMC to verify borderline cases before fine issuance.
- Public awareness campaigns on how to contest erroneous e‑challans.
- Periodic audits of AI performance and transparent reporting of accuracy metrics.
While AI can reduce manual harassment and speed up enforcement, the Bengaluru experience shows that technology must be complemented by robust human checks and clear redressal pathways to maintain public trust.