# Can 80,000 Cameras Think as One? Inside Gujarat Police’s Mega AI Hackathon > Source: > Published: 2026-08-17 09:06:47+00:00 # Can 80,000 Cameras Think as One? Inside Gujarat Police’s Mega AI Hackathon Gujarat has more than 80,000 police and government CCTV cameras. The problem is that many of them do not speak the same technological language. Different departments use different vendors, networks, software and video-management systems. One camera may spot a suspicious vehicle, another may record it five kilometres away, but the systems may struggle to connect those sightings quickly. Gujarat Police now wants coders to make those cameras work as one. The Gujarat Police Innovation Challenge 2026, billed as India’s largest hackathon devoted to CCTV integration and AI-based video analytics, will bring students, startups, researchers and technology giants into the same contest. Their mission: build AI tools that can scan live feeds, recognise suspicious activity, track vehicles and alert police before the trail goes cold. In other words, Gujarat has the eyes. The hackathon must build the brain. ## The Age of Anger 31 Jul 2026 - Vol 05 | Issue 31 A raging generation makes the government relent. What's next? [Read Now The Age of Anger](/magazine/the-age-of-anger) ## What exactly is Gujarat Police trying to build? A unified surveillance network connecting more than 80,000 CCTV cameras across the state. Those cameras currently operate across different government departments and technological systems. Some may run on older infrastructure, while others use proprietary software that does not easily exchange information with rival platforms. The hackathon will hunt for a scalable layer capable of pulling these fragmented feeds together. If it works, police control rooms could search, monitor and analyse footage across the network instead of jumping between disconnected systems. ## Why does Gujarat need a hackathon for that? Because joining 80,000 cameras is not as simple as plugging them into one very large television screen. Each camera may generate video in a different format, operate at a different resolution and connect through a different network. Vendors may use their own software protocols, while departments may store footage for different periods and label it differently. The winning solution must make these technological islands communicate without demanding that Gujarat replace every camera or rebuild its surveillance infrastructure from scratch. That is the hackathon’s real puzzle: make old and new machines cooperate at state scale. ## What will the AI actually do? The proposed tools will be tested on their ability to detect people, vehicles and unusual activity, then generate alerts in real time. The challenge will include automatic number-plate recognition, vehicle tracking, watchlist matching and cross-camera searches. A police team looking for a particular car, for instance, could theoretically enter its number or distinguishing features and trace its movement across multiple locations. Cross-camera search could also help investigators find the same person or vehicle across hours of footage without forcing officers to watch every recording manually. The AI would not replace the cameras. It would sift through what they see. ## Can a CCTV camera really recognise “suspicious” behaviour? Not by itself. A camera records pixels. Software assigns meaning to them. AI systems can be trained to flag patterns such as a vehicle stopping in a prohibited zone, an unattended object, movement inside a restricted area or a person matching an authorised watchlist. But “unusual” does not automatically mean “criminal”. Someone running may be fleeing a crime, chasing a bus or late for work. A bag left unattended may contain a threat or someone’s lunch. That is why alerts require human verification. If the software treats every anomaly as guilt, smart surveillance can quickly become automated suspicion. ## What makes this hackathon different? Participants will reportedly develop and test their tools using live CCTV feeds and real-world policing scenarios. Gujarat Police says this will be the first Indian competition of its kind to put contestants’ solutions through such an operational test. Instead of demonstrating software on a carefully prepared dataset, teams will have to prove that it works amid poor lighting, crowded roads, obstructed views, changing weather and unpredictable human behaviour. A prototype can perform beautifully in a laboratory. A rain-smeared camera at a congested junction is a different examination. ## Who can enter? The Open Innovation Challenge will have two categories. One will cater to students and small and medium-sized startups. The other will bring in larger startups and established technology companies. The structure gives younger teams a chance to compete without being immediately overwhelmed by companies with large engineering departments and deeper pockets. Six of the best-performing teams from the opening stage will advance to the finale, where they will demonstrate their systems in a live production environment. ## What do the winners get? The competition carries total cash prizes of ₹37 lakh. The more valuable reward, however, could be the opportunity to test a product across one of India’s largest proposed integrated surveillance networks. A solution that survives Gujarat’s 80,000-camera challenge could potentially be adapted by police forces, municipal bodies and transport systems elsewhere. For startups, that makes the hackathon more than a coding contest. It could become a showroom. ## When does it begin? The hackathon is expected to start in September 2026. Gujarat Police has published participation details, eligibility conditions and competition rules on its official portal. The initiative is being organised with technology support from i-Hub Gujarat, while DA-IICT and the National Forensic Sciences University are serving as knowledge partners. ## Will AI allow police to watch everyone all the time? Technically, an integrated system would greatly expand the police’s ability to search and connect footage. Whether that becomes indiscriminate surveillance depends on the rules governing its use. The announcement explains what the technology should detect, but does not provide extensive details about data retention, access controls, audit trails, independent oversight or how citizens could challenge a false match. Those questions matter. A centralised network can help police reconstruct a crime quickly, locate a missing person or track a fleeing vehicle. The same capacity, if poorly controlled, can also map an innocent person’s movements across cities. A system this powerful needs boundaries as sophisticated as its algorithms. ## How accurate is AI surveillance? Accuracy varies sharply with image quality, camera angle, lighting, crowd density and the data used to train the system. Number-plate software may misread dirty, damaged or partially hidden plates. A person can be incorrectly matched because of poor footage or resemblance to someone on a watchlist. Systems tested on narrow datasets may also perform unevenly across age groups, genders and skin tones. At the scale of 80,000 cameras, even a low error rate can produce a large number of false alerts. Human review, documented verification and regular accuracy audits will therefore be essential. ## Could this speed up policing? Yes, particularly when investigators already know what they are seeking. After a robbery, police could search for a getaway vehicle across cameras instead of manually collecting footage from multiple departments. During an emergency, the network could help locate blocked roads, detect abnormal crowd movement or track a vehicle across districts. It could turn hours of video hunting into minutes of targeted searching. Faster is useful, however, only when the result is reliable. ## What is the hackathon’s biggest challenge? It is not teaching 80,000 cameras to see. They already do that. The challenge is teaching different systems to share what they see, getting AI to separate genuine threats from ordinary behaviour and ensuring police can act on alerts without blindly trusting an algorithm. If Gujarat succeeds, it could build a powerful model for technology-driven policing. If safeguards lag behind capability, it could also build a machine that watches far more effectively than it understands. *(With inputs from ANI)*