A small control room tucked behind Delhi’s Daryaganj Police Station has three computers, an 11-member team, and a steady stream of beeping alerts that animate the space. On the walls, posters read “Say Help”, the emergency app now wired into the Central district’s policing system.
Say Help is a real-time safety platform that lets users trigger distress alerts through a tap or up to five custom voice phrases. Once activated, it silently switches on the phone’s camera and microphone, captures GPS coordinates, and sends a live feed to emergency contacts and the police. Available on Android and iOS, it shares continuous video, audio, and location tracking until the alert is closed.
The app has been downloaded over 5000 times in India. And on an average, 40 pings reach the app control room every day.
Integrated with the Central district police in September 2025, the system mostly receives test alerts, people experimenting after hearing about the app. But within these test pings are the alerts that matter.
One such alert, sent in the intervening night of October 30–31, helped Delhi Police crack a trafficking racket. A woman trapped inside a spa on DBG Road used the app to send an SOS. Intelligence already suggested the premises—Galaxy Spa, near Today Hotel—was being used as a front for organised trafficking.
A decoy officer entered with marked currency and confirmed the activity through a pre-arranged missed WhatsApp call. Around 7:25 pm, a police team raided the location, rescuing six women. Police recovered the marked cash and sealed packets of condoms from the site.
Criminologist Ntasha Bhardwaj said that while the app may have helped police rescue six women that night, it also reveals a broader pattern in how safety technologies tend to be positioned. Reflecting on the case, she argued that tools like Say Help can potentially frame complex issues such as trafficking as problems that can be solved primarily through rapid police response.
“This framing reduces structural harms to matters of law and order,” she said, explaining that when safety is designed around surveillance and enforcement, deeper questions of survivor justice, long-term support, and prevention tend to fade into the background.
Even with real-time alerts, she added, “the end point is still the police—and systemic delays may not change.”
How the App Was Born
Say Help was built by US-based tech professional Mahesh Salgaonkar in 2023. The trigger, he recalled, was a case in the US involving two teenage girls who were raped and murdered. One had taken a photo of the perpetrator, but privacy laws stopped police from accessing her phone for years.
An investigating officer, also a friend, asked if such barriers could be bypassed in emergencies. “Could there be a tool that gives police live video, audio, and location?” he had asked Salgaonkar.
“It took a couple of years,” Salgaonkar told Decode. “By 2023, the app was ready.”
He later pitched it to DCP Nidhin Valsan during a meeting in Goa. Valsan expressed interest, and once he took charge of the Central district, he reached out. Integration was completed in September 2025.
The developer is now in talks with police departments in Uttar Pradesh, Gujarat, Punjab, and Madhya Pradesh. His team often visits trade fairs, colleges, and public events to demonstrate the app.
What A ‘Help’ Triggers
Once the app is installed, users register and add trusted contacts. The app requests access to the camera and microphone, which can be granted “only this time” or “while using the app”—and nudges users toward the latter. Once launched, it can keep running quietly in the background.
Alerts can be triggered by tapping ‘Send alert’ inside the app or by using the voice command “Help, help, help”. The voice trigger works only after users tap ‘Activate’ once; after that, they can close the app and use the voice command as long as the phone’s microphone can pick it up.
The Voice Trigger Functions Only After Users Tap ‘Activate’
Users can set up to five custom trigger phrases in any language. Each phrase is registered by repeating it three times; the system accepts it if it recognises the phrase correctly at least twice.
When triggered, the phone begins streaming without any visible sign. A link is sent instantly to emergency contacts and the control room. The link opens into a dashboard showing the user’s details, the live video feed, and a real-time map that tracks movement and speed until the alert is closed.
Say Help is being used in 18 countries including Mexico, India, Chile, Peru, the UAE, Kuwait, and Colombia.
Nikhil Bharadwaj, a Delhi Police constable overseeing the Say Help project, explained how they separate real emergencies from tests inside the control room. The first few moments of the live feed are crucial: the team watches the user’s expression, tone, surroundings, and background noise.
“With experience, you start recognising what looks staged and what looks real,” he said. Alerts from women and senior citizens are prioritised. Even when unsure, he added, “we still send it. Better to alert the police than miss a genuine case.”
The Roadblocks
Accidental triggers are frequent, often more than 10 a day. In fact, only one or two alerts usually warrant being forwarded to the police control room.
One recent incident involved a 72-year-old woman whose alert seemed urgent. The police reached her house in 19 minutes, only to discover that her daughter had pressed the button to test the app.
When an alert appears genuine, the team manually identifies the nearest police station and shares a link mirroring the one sent to emergency contacts. They can track police movement in real time as officers head towards the user.
The planned AI features for the app have hit some roadblocks—some challenges are technical; others relate to the complexity of training AI for messy real-world use.
A major setback came from Delhi Police’s geo-fencing design. Ideally, police boundaries are mapped precisely. But in Delhi, these borders are thick 100-metre buffer zones, which the AI struggled to interpret.
“Most crimes happen exactly in these border areas,” Salgaonkar said. “But the AI couldn’t map them confidently.” As a result, alerts couldn’t be routed automatically, and the team had to switch to manual tracking.
Another planned feature is crime-type detection, where AI analyses video and audio to infer whether an incident involves harassment, assault, domestic violence, or something else. But this requires extensive training across accents, dialects, and noisy environments. “It will take time,” Salgaonkar said.
“Bridging the Civilian–Police Gap”
According to Bharadwaj, Say Help is designed to reduce response time. While emergency number 112 functions “like a call centre,” he said, Say Help works on a trigger mechanism that nudges police into immediate action.
The app has already been used in several non-trafficking cases: stopping a digital-arrest scam targeting a senior citizen, intervening when college students were harassed on GB Road, and alerting UP Police to a domestic violence situation in Noida.
Most recently, the app helped police respond quickly to a case in Dwarka, where a woman living alone alleged that her landlord disconnected the power, forced his way in, and began harassing her. An alert sent through the app brought officers to her door in time.
The platform also allows bystanders to report crimes without revealing their identity. They don’t need to appear before the police or file a complaint themselves, the constable said.
The Privacy and Misuse Concerns
But the tool also raises red flags.
Because the app can run in the background and relies on camera and microphone access once the user taps ‘Activate’, those who are less digitally experienced may unknowingly leave it in a state where it continues listening for voice triggers even when they don’t intend it to. The app can also be invoked during mutual fights or disputes, potentially prompting unnecessary police action or recording bystanders without their consent.
Salgaonkar maintained that the platform includes checks against misuse—his team reviews each alert and shuts down those that appear frivolous, the video feed can’t be downloaded, and police can access footage only through a Section 94 BNSS request. “Even my staff can’t access history once an alert is closed,” he said. Only one authorised person in Delhi Police can search a specific alert.
A Delhi Police Personnel Monitors Incoming Alerts On The Say Help Dashboard
“We have tried to think of all the scenarios to prevent any mishap. So far the system has performed well,” he added.
Criminologist Ntasha Bhardwaj, however, cautioned that covert recording collapses the boundary between protection and surveillance.
“There’s no clear consent framework—for bystanders, for those being recorded, or even for the user,” she said. She added that hidden surveillance has historically been deployed disproportionately against poor, migrant, Dalit, Muslim and informal workers.
“It expands police power without expanding accountability,” the criminologist said. “Safety built on invisibility and covert monitoring isn’t empowerment—it’s conditional protection at the cost of privacy.”