AI for Emergency Response 112 ERSS: Fixing India's Unified Emergency Number

Executive Summary

India's 112 Emergency Response Support System is the unified emergency number covering all 36 States and Union Territories — and it is being overwhelmed by non-emergency traffic at a scale that defies manual management. Telangana receives approximately 16 lakh daily calls through its integrated 112 network, of which only 0.28% represent genuine emergencies requiring police or ambulance intervention. Punjab's 2025 Comptroller and Auditor General (CAG) report found that 86% of mandated emergency response vehicles were missing, no Standard Operating Procedures existed after nine years of operation, and calls took more than 30 minutes to reach incident sites in 52% of cases. At the national level, the system processes an estimated 16–19 lakh calls per day across 36 Emergency Response Centres and 632 District Coordination Centres.

Executive Callout Telangana's 112 network receives approximately 16 lakh daily calls; only 0.28% are genuine emergencies. Punjab's CAG 2025 audit documented an 86% shortfall in emergency response vehicles and a 28–31% rate of calls that took over 5 minutes just to forward to dispatchers. Haryana deployed AI-based auto-dispatch in July 2025 and cut average police response time from 12 minutes 4 seconds to 7 minutes 3 seconds, achieving 92.6% caller satisfaction. C-DAC holds the Rs 531.24 crore ERSS Phase II contract through March 2026, with a NextGen roadmap that explicitly includes AI integration. The market is open, the model is proven, and the procurement pathway runs through a partnership with C-DAC. (Aisewak Government Helpline Report, 2026, citing CAG Report No. 7/2025 Punjab, KPIAS Academy analysis Telangana, C-DAC ERSS Phase II documentation.)

The single most actionable insight for government leaders reading this article: AI is not a future consideration for 112 ERSS. It is a present solution to a documented crisis. Haryana has already demonstrated a 42% reduction in police response time using AI-based auto-dispatch. The technology is available, the procurement pathway through C-DAC is established, and the cost of inaction — measured in lives lost during delayed emergency response — is no longer defensible.

Introduction

When a citizen calls 112 in an emergency, every second matters. The difference between a dispatcher answering in 90 seconds versus 5 minutes is not an operational statistic — it is the difference between an accident victim receiving timely medical attention and one who does not. Yet India's 112 system, despite years of investment and infrastructure build-out, is spending the vast majority of its dispatcher capacity on calls that are blank dials, pocket calls, prank calls, and wrong numbers.

The 112 ERSS was designed as India's answer to the United States' 911 system — a single number that routes voice calls, SMS messages, SOS signals, and even IoT emergency alerts to the nearest Emergency Response Centre, then dispatches police, fire, or ambulance as required. C-DAC Thiruvananthapuram serves as the Total Service Provider for ERSS Phase II, deploying a 130-engineer team on a Rs 531.24 crore project. The system now supports ten incoming channels: voice call, SMS, SOS, email, web request, chatbot, media crawler, IoT signals, WhatsApp, and external signals. (Aisewak Government Helpline Report, 2026, citing C-DAC ERSS Phase II documentation.)

The infrastructure exists. The problem is not investment — it is signal-to-noise ratio. When 99.5% of calls are spam, the system's human capacity is consumed by irrelevant traffic, and the 0.5% of genuine emergencies compete for dispatcher attention with millions of accidental dials.

Current Challenges: A System Overwhelmed by Its Own Volume

The Spam Crisis

The scale of non-emergency traffic on 112 is not disputed. It is documented in official audit reports and state-level analyses that government officials cannot dismiss.

State	Problem Documented	Source
Telangana	99.5% of ~16 lakh daily calls are spam; only ~4,500 genuine emergencies per day	KPIAS Academy analysis (Aisewak Government Helpline Report, 2026)
Delhi	61.27% blank calls; 96% of 112 calls rejected by IVRS (February 2020)	CAG Report No. 15 of 2020
Punjab	86% vehicle shortfall; 52% of calls take >30 min to reach incident sites	CAG Report No. 7 of 2025
Punjab	28–31% of calls take >5 minutes just to forward to dispatchers	CAG Report No. 7 of 2025
Punjab	Calls forwarded within 2 minutes: only 2–4%	CAG Report No. 7 of 2025

Source: Aisewak Government Helpline Report, 2026.

Delhi's CAG data reveals a pattern that predates the current system: even with an Interactive Voice Response System (IVRS) filter, 61% of calls were blank — meaning the caller hung up without speaking. These blank calls occupy dispatcher lines, trigger alert protocols, and consume management bandwidth. The IVRS rejected 96% of incoming calls in February 2020, suggesting the filtering mechanism itself was calibrated so aggressively that it was likely rejecting genuine emergencies along with spam.

The Telangana data is the most extreme: with 16 lakh daily calls and only 4,500 genuine emergencies, a human dispatcher working 8-hour shifts could theoretically handle every genuine emergency call in the state. But when those 4,500 genuine calls are buried under 15.96 lakh irrelevant ones, the dispatcher's entire capacity is consumed by noise — and genuine emergencies wait.

The Infrastructure Gap: Punjab's CAG Findings

The Punjab CAG Report No. 7 of 2025 is the most comprehensive recent audit of ERSS performance, and its findings are severe enough to warrant direct quotation in any policy briefing.

Against a requirement for 1,866 emergency response vehicles (ERVs) and two-wheelers, Punjab deployed only 258. That is an 86% shortfall. The two-wheeler shortfall was 98%. Nine years into ERSS operation, no Standard Operating Procedures had been developed. Of the 24 State Advisory Committee meetings required under ERSS guidelines, only one had been held.

The central government disbursed funds to Punjab with delays ranging from one to five years, accumulating Rs 1.24 crore in interest that was never passed on. The consequence was predictable: in 52% of documented emergency responses, vehicles took more than 30 minutes to reach incident sites. For reference, the international benchmark for emergency response in urban areas is under 8 minutes.

These failures are not attributable to technology gaps alone. They reflect a combination of procurement delays, workforce management failures, and monitoring gaps that technology can address only partially. But the call-handling component — the 28–31% of calls that took over 5 minutes just to forward to a dispatcher — is precisely the problem that AI voice technology solves directly.

Why Traditional Dispatch Cannot Scale

The mathematics of manual dispatch are unforgiving. A human dispatcher can handle one call at a time. During a peak-hour surge, a festive season, or a natural disaster, call volume can multiply 3–5 times within hours. Human capacity cannot scale proportionally.

The consequences are measurable. When dispatchers are occupied handling spam calls, genuine emergencies wait. When call volumes spike, answer rates fall. When answer rates fall, citizens in distress redial — adding repeat calls to an already-overloaded queue. This feedback loop is documented across every high-volume government helpline in India. (Aisewak Government Helpline Report, 2026, citing CAG data across multiple states.)

The 112 system's multi-channel design — voice, SMS, SOS, email, WhatsApp, IoT — compounds the challenge. A unified emergency number that accepts ten input channels without AI-powered triage requires a proportional increase in human review capacity for each new channel. Adding WhatsApp as an emergency channel without AI filtering does not improve response; it adds another stream of noise to process manually.

How AI Voice Technology Solves the 112 Problem

The AI solution for 112 ERSS is not a single product. It is a three-layer architecture that addresses the spam problem, the dispatch problem, and the multilingual access problem in an integrated sequence.

Layer 1: Real-Time Spam Classification

An AI voice agent that answers every 112 call within two seconds and classifies intent within 15 seconds is technically feasible with current generation large language model (LLM) plus speech-to-text infrastructure. The classification categories are well-defined:

Category	Description	AI Action
Genuine emergency	Caller describes a specific incident requiring police, fire, or ambulance	Immediate transfer to human dispatcher with context
Blank call / silent call	No voice detected within 5 seconds	Auto-flag for callback; log caller ID
Pocket dial	Caller unaware they dialled 112	Auto-resolve with a brief message
Prank call	Deliberate false emergency report	Log, apply repeat-caller flag, transfer to duty officer
Enquiry / wrong number	Non-emergency information query	Redirect to 1800 helpline or relevant department
Repeat spam caller	Previously identified spam number	Accelerated classification; escalate if pattern changes

The classification accuracy target is achievable: 90%+ spam detection with a false positive rate (genuine calls misclassified as spam) below 2%. The 2% false positive threshold is conservative — it means that for every 100 calls the AI classifies as spam, no more than 2 genuine emergencies are misclassified. Given Telangana's ratio of 4,500 genuine calls in 16 lakh total, a 2% false positive rate would misdirect 319 genuine calls per day — which is why the architecture requires a safety net: all borderline classifications route to a human dispatcher rather than auto-dropping.

Layer 2: Integration with C-DAC's NextGen ERSS

C-DAC's NextGen ERSS (NG-ERSS V2.0) roadmap already includes AI chatbot integration, speech-to-text, and intelligent routing. The AI spam filter does not require replacing C-DAC's architecture — it plugs into the existing NG112 API as a pre-dispatch classification layer.

This is a critical point for procurement framing: the proposition to C-DAC and to state governments is not "replace your existing system" but "add a classification layer that makes your existing investment more effective." AI voice classification reduces the load on human dispatchers, improves response times for genuine emergencies, and generates data analytics that C-DAC's PMIS (Project Management Information System) cannot currently produce.

Karnataka's new government-owned Command and Control Centre runs on C-DAC NG-ERSS V2.0, with a 50-seat call centre integrating 108, 104, 112, 181, 1098, Tele-MANAS, and eSanjeevani. (Aisewak Government Helpline Report, 2026, citing C-DAC Karnataka deployment data.) This integrated architecture is the template: AI classification at the front end, human judgment at the dispatch layer, and unified data at the analytics layer.

Layer 3: Multilingual Emergency Response

India's 112 system has a structural language gap. Hindi and English cover the primary response interface, but genuine emergencies are reported in 22 scheduled languages and hundreds of dialects. A Tamil-speaking caller in distress who cannot communicate in Hindi faces a barrier precisely when they are most vulnerable.

Bhashini's production-ready multilingual voice infrastructure — supporting 22 languages in voice recognition and processing 15 million-plus AI inferences daily across government platforms — provides the solution. (Aisewak Government Helpline Report, 2026, citing MeitY Bhashini data.) An AI voice agent using Bhashini's CONVERSE API can conduct the classification conversation in the caller's language, extract the essential emergency information (location, nature of incident, number of persons involved), and pass structured data to the dispatcher in the dispatcher's working language.

For a state like Telangana, where Telugu-speaking callers form the primary user base but the dispatch system operates partly in Hindi and English, this multilingual bridge has immediate operational value.

The Haryana Model: Proof That AI Works in Emergency Response

The most credible argument for AI in 112 ERSS is not theoretical — it is the documented outcome from Haryana.

In July 2025, Haryana deployed AI-based auto-dispatch for its 112 system. The results, reported by the state government and verified independently, are measurable:

Average police response time fell from 12 minutes 4 seconds to 7 minutes 3 seconds — a 42% reduction.
Caller satisfaction reached 92.6%, earning national recognition from the Ministry of Home Affairs.
The system demonstrated that AI dispatch can make time-critical routing decisions — which ERV is closest, which officer is available, which route avoids current traffic — faster and more accurately than a dispatcher managing multiple calls simultaneously.

(Aisewak Government Helpline Report, 2026, citing Haryana 112 state government data.)

The Haryana model has three characteristics that make it directly replicable: it operates on the standard C-DAC NG-ERSS platform, it uses Bhashini-compatible Hindi voice processing, and its KPI metrics (response time in seconds, satisfaction percentage) are quantifiable for tender documentation in other states.

States with documented response time failures — Punjab (52% exceeding 30 minutes), Delhi (61% blank calls consuming dispatcher capacity) — have both the performance gap and the political pressure to replicate Haryana's model.

Implementation Roadmap for Government Leaders

Government leaders evaluating AI for 112 ERSS should structure implementation across three phases, aligned with C-DAC's Phase II timeline and state budget cycles.

Phase 1: Pilot (Days 1–90) — Spam Classification in One State

The recommended pilot location is Haryana, for a specific reason: the state has established baseline data (12 min 4 sec baseline response time, 92.6% post-AI satisfaction) and leadership with demonstrated openness to technology innovation. A pilot that replicates and validates the AI dispatch model on an expanded scope — adding spam classification to the existing auto-dispatch — provides the most credible proof-of-concept.

Pilot parameters:

Scope: Real-time spam classification on 50,000+ daily calls
Languages: Hindi, Haryanvi dialect, English
KPIs: Spam classification accuracy >90%; genuine emergency detection rate >99%; average call classification time <15 seconds; false positive rate <2%
Cost: Rs 25–35 lakh for pilot infrastructure
Timeline: 30 days to go-live from MoU signing, 60 days of operation, 30 days of data analysis

The pilot success metric is binary: demonstrate that over 80% of spam calls are auto-filtered without a single genuine emergency call being lost to the system.

Phase 2: State-Scale Deployment (Months 4–12)

After pilot validation, state-scale deployment covers all Emergency Response Centres and District Coordination Centres in the pilot state. The integration requirements at this phase include:

C-DAC NG112 API connection for dispatch integration
Bhashini voice model deployment for regional language support
State GIS and AVLT (Automatic Vehicle Location Tracking) system integration for real-time dispatch optimisation
Analytics dashboard integration with state PMIS

Cost estimate for full state deployment: Rs 2–3 crore, based on the Haryana model and comparable pilot data from Karnataka.

Phase 3: National Replication (Year 2 Onwards)

C-DAC's role as Total Service Provider creates a natural replication mechanism: once the AI classification layer is integrated into the NG-ERSS platform, it is available to all 36 states and UTs that use the C-DAC system. National replication does not require 36 separate procurement processes — it requires a single integration agreement with C-DAC, followed by state-level customisation (languages, dispatch protocols, SLA targets).

States with active procurement windows — Telangana (explicit AI deployment interest for spam filtering), Punjab (post-CAG Rs 178 crore overhaul with AI call handling planned) — represent the second wave of deployment.

Expected Impact: Before-vs-After Comparison

Metric	Current State (National Average)	Expected Post-AI Outcome	Evidence Basis
Spam call auto-filtering	0% (manual handling)	>90% automated	AI classification benchmarks; Telangana data
Average call classification time	>5 minutes (Punjab: 28–31% of calls)	<15 seconds	AI speech-to-text latency benchmarks
Genuine emergency response time	12+ minutes (urban); 30+ min (Punjab 52%)	<8 minutes	Haryana model: 12:04 → 7:03
Dispatcher capacity freed for genuine calls	~0.5% of calls are genuine (Telangana)	All dispatcher capacity on genuine calls	Direct consequence of spam filtering
Caller satisfaction	Not systematically measured	>90%	Haryana model: 92.6%
Languages supported	Hindi + English primary	8+ languages via Bhashini	Bhashini IndicTrans2 production capability
Operating hours	24/7 (human capacity constrained)	24/7 (AI handles surge without staffing costs)	Structural benefit of AI scale

Source: Aisewak Government Helpline Report, 2026, citing CAG Punjab 2025, Haryana 112 state data, C-DAC ERSS documentation.

The return-on-investment calculation for government decision-makers rests on a single insight: dispatcher time currently wasted on 99.5% non-emergency calls is the most expensive misallocation in India's emergency response system. If AI classification frees even 50% of that wasted capacity, the number of genuine emergencies that receive timely human response effectively doubles — without a single additional dispatcher being hired.

Risks and Mitigation

Any deployment of AI in life-safety systems requires an honest risk assessment. Three risks are material for 112 ERSS.

Risk 1: False positives — genuine emergencies classified as spam. This is the most consequential risk. A silent call from a domestic violence victim could be misclassified as a blank call. A caller in shock who cannot speak coherently could be misclassified as a prank. The mitigation is a tiered classification protocol: high-confidence spam calls are logged and dropped; borderline classifications (any call where the AI's confidence is below 85%) route to a human dispatcher. The 2% false positive target requires ongoing model monitoring and quarterly recalibration against dispatcher feedback data.

Risk 2: Procurement complexity through C-DAC. No major emergency helpline procurement has bypassed C-DAC in the past five years. A direct tender approach without C-DAC partnership faces 18–36 month cycles. The mitigation is partnership-first: position AI classification as a complementary layer to NG-ERSS rather than a competitive product, and pursue integration through C-DAC's Phase II work-order mechanism.

Risk 3: Multilingual quality gaps. Bhashini's voice models for regional languages vary in quality. Hindi and English are production-grade; some regional dialects remain in active development. The mitigation is phased language deployment — launch with languages where Bhashini achieves >95% word error rate benchmarks, and add additional languages as model quality improves.

Future Outlook: From Emergency Response to Intelligent Governance

The 112 ERSS modernisation case illustrates a broader principle in Indian governance: AI's first and most defensible use case in citizen services is not intelligence augmentation but noise reduction. Before AI can help governments make smarter decisions, it must help emergency dispatchers focus on genuine emergencies rather than pocket dials.

The natural next step after spam classification is intelligent dispatch — AI that not only filters calls but recommends the optimal emergency response (which vehicle, which route, which hospital) based on real-time GIS data, traffic, and resource availability. Haryana's July 2025 deployment is the proof of concept. The next five states that replicate it will generate the data for a national standard.

Looking further ahead, the multi-channel design of NG-ERSS — voice, SMS, SOS, WhatsApp, IoT signals — creates the architecture for predictive emergency response: AI that identifies emerging cluster patterns (multiple calls from the same location within minutes) before a dispatcher recognises the pattern manually. This moves 112 from reactive to proactive — from answering emergencies to anticipating them.

For state government leaders, the strategic question is not whether to deploy AI in 112 but in what sequence. The Haryana-first approach — start with a proven model, measure rigorously, replicate — is the lowest-risk path to transformational impact.

Key Takeaways

India's 112 ERSS receives 16–19 lakh daily calls nationally, of which 99.5% in documented states are spam or non-emergency. The signal-to-noise ratio is the primary barrier to effective emergency response.
Punjab's CAG 2025 report documented an 86% emergency response vehicle shortfall, no SOPs after nine years, and a 52% rate of response times exceeding 30 minutes — failures that AI call classification can partially address by ensuring every genuine emergency reaches a dispatcher without delay.
Haryana's July 2025 AI-based auto-dispatch reduced police response time from 12:04 to 7:03 minutes and achieved 92.6% caller satisfaction — a replicable model available to every state using C-DAC's NG-ERSS platform.
C-DAC's NextGen ERSS already includes AI chatbot integration in its roadmap, creating a partnership-first procurement path that bypasses the 18–36 month direct tender cycle.
The recommended pilot — 30 days in Haryana at Rs 25–35 lakh — validates AI spam classification before state-scale investment, with a single measurable success criterion: >80% spam auto-filtered with <2% false positive rate.
Multilingual capability through Bhashini extends genuine emergency access to Tamil, Telugu, Kannada, Bengali, and Punjabi-speaking callers who currently face a language barrier in India's primary emergency number.

Conclusion

India's 112 Emergency Response Support System is one of the country's most consequential citizen-facing infrastructure investments. The Rs 531.24 crore C-DAC Phase II contract and the decade of state-level build-out represent a genuine commitment to unified emergency response. The gap between that commitment and current performance — documented in CAG reports, state-level audits, and the Telangana spam data — is not a reason to question the investment. It is a reason to complete it with the one capability that human-only systems cannot provide: real-time, 24/7, multilingual classification of 16–19 lakh daily calls, at the speed and scale that AI makes possible.

The Haryana model has answered the question of whether AI works in Indian emergency response. The remaining question is how quickly the other 35 states and Union Territories will act on that evidence.

Government leaders exploring AI-powered citizen engagement can begin with a focused pilot in one department or constituency to validate impact before scaling statewide. Aisewak helps public institutions deploy multilingual Voice AI solutions designed specifically for Indian governance.

FAQ

Q1: What is 112 ERSS and how does it differ from earlier emergency numbers like 100, 101, and 108?

112 ERSS (Emergency Response Support System) is India's unified emergency number, launched to consolidate police (100), fire (101), ambulance (108/102), and women's helpline (181) under a single, technology-enabled response system. Unlike the earlier siloed numbers, 112 routes calls to a central Emergency Response Centre that dispatches the appropriate service. C-DAC Thiruvananthapuram serves as the Total Service Provider for ERSS Phase II, managing the system across all 36 States and UTs.

Q2: Why does India's 112 helpline receive so many spam calls?

The combination of a single universally advertised emergency number, widespread smartphone ownership, and low awareness of consequences for false calls creates a high volume of accidental and prank calls. Telangana data shows approximately 99.5% of 16 lakh daily calls are spam or non-emergency; Delhi's 2020 CAG report documented 61.27% blank calls. Repeat offenders are a concentrated problem — Delhi data showed 20% of blank calls came from just 15 repeat callers.

Q3: How does AI classify emergency calls without risking genuine calls being dropped?

AI classification uses a tiered confidence approach. High-confidence spam (blank calls, known repeat spammers, pocket dials) is auto-resolved. Borderline classifications — any call where the AI's confidence is below a defined threshold — route directly to a human dispatcher. This ensures that the false positive rate (genuine calls misclassified as spam) remains below 2%, and that no genuine emergency is lost without human review.

Q4: What was Haryana's AI-based 112 deployment and what did it achieve?

Haryana deployed AI-based auto-dispatch for its 112 system in July 2025. The AI optimised ERV dispatch in real time based on vehicle location, route, and incident type. The outcome was a reduction in average police response time from 12 minutes 4 seconds to 7 minutes 3 seconds — a 42% improvement — and a caller satisfaction rate of 92.6%, earning national recognition from MHA. This is the most credible proof point for AI in Indian emergency response currently available.

Q5: What is C-DAC's role in 112 ERSS and why does it matter for AI procurement?

C-DAC Thiruvananthapuram is the Total Service Provider for ERSS Phase II, a Rs 531.24 crore project. Its NG-ERSS V2.0 platform is the technological backbone of 112 across states, including Karnataka's new government-owned Command and Control Centre. No major emergency helpline technology procurement has bypassed C-DAC in the past five years. For AI vendors, this means partnership with C-DAC is the prerequisite for meaningful market access — direct tenders without C-DAC involvement face 18–36 month cycles.

Q6: Which states are most urgently in need of AI deployment for 112?

Based on documented failure rates and active procurement interest: Telangana (99.5% spam — highest noise ratio nationally, state police have publicly discussed AI deployment), Punjab (post-CAG Rs 178 crore overhaul with AI call handling planned), and Delhi (61% blank calls, 96% IVRS rejection rate). Haryana is the model state for replication rather than a priority for intervention, having already deployed AI dispatch successfully.

Q7: How does multilingual AI support improve 112 response for non-Hindi speakers?

The current 112 interface operates primarily in Hindi and English. For Tamil, Telugu, Kannada, Bengali, and Punjabi-speaking callers — who collectively represent hundreds of millions of citizens — a language barrier in an emergency can delay or prevent effective help. Bhashini's IndicTrans2, supporting 22 languages in voice recognition, enables AI voice agents to conduct the classification and triage conversation in the caller's language and pass structured information to the dispatcher in the dispatcher's working language. This removes the language barrier without requiring multilingual human dispatchers.

Q8: What does a 30-day pilot for 112 AI actually involve and what does it cost?

A pilot on the Haryana model would run AI spam classification on 50,000+ daily calls for 30 days. Scope includes real-time intent classification, integration with the existing C-DAC NG112 API, and daily accuracy reporting. Languages would cover Hindi, Haryanvi dialect, and English. KPIs are spam classification accuracy >90%, genuine emergency detection >99%, and call classification time <15 seconds. Estimated cost is Rs 25–35 lakh for pilot infrastructure and 30-day operations, rising to Rs 2–3 crore for full state deployment.

Q9: How does AI dispatch reduce emergency response times, not just filter spam calls?

AI dispatch addresses a separate challenge: once a genuine emergency is identified, the AI system uses real-time GIS data, vehicle location (AVLT tracking), traffic conditions, and incident type to recommend the optimal ERV dispatch rather than relying on the dispatcher's manual assessment. In Haryana's model, this optimisation reduced average response time by over 5 minutes — time that is directly consequential in cardiac emergencies, fire containment, and violent incident response.

Q10: What KPIs should government officials demand from any AI vendor proposing 112 modernisation?

The minimum KPI framework should include: spam classification accuracy (target >90%); genuine emergency false positive rate (target <2%); average call classification time (target <15 seconds); percentage of genuine emergencies reaching human dispatcher within 30 seconds (target >99%); average emergency response time before and after AI deployment (measured in seconds); and caller satisfaction rate via post-call IVR survey (target >85%). Any vendor unable to commit to these metrics in a pilot SLA should not be considered for production deployment.

Q11: Is AI deployment in 112 compliant with India's DPDP Act 2023?

The Digital Personal Data Protection Act 2023 requires explicit consent for processing personal data. Emergency calls are addressed under a legitimate interest exemption that covers public safety functions. However, AI systems deployed in 112 should minimise data retention: call classification data should be retained only for the minimum period required for quality auditing, and personally identifiable information from spam calls should be purged within 90 days. Government procurement documents for 112 AI should explicitly reference DPDP compliance requirements. For a detailed treatment, see our analysis of DPDP Act compliance for government voice AI.

Schema Markup Suggestions

Primary Schema Types:

Article — for the main post body
FAQPage — for the FAQ section (each Q&A pair as Question + AcceptedAnswer)
GovernmentService — for the 112 ERSS service description

Key Fields:

Article.headline: "AI for Emergency Response 112 ERSS: Fixing India's Unified Emergency Number"
Article.author: Aisewak Research Team
Article.datePublished: 2026-07-24
GovernmentService.name: "112 Emergency Response Support System (ERSS)"
GovernmentService.provider: C-DAC Thiruvananthapuram / Ministry of Home Affairs
GovernmentService.serviceType: "Emergency Response"
GovernmentService.areaServed: "India"

Suggested External References

CAG Report No. 7 of 2025 (Punjab): Performance Audit of Emergency Response Support System 112
CAG Report No. 15 of 2020 (Delhi): Performance Audit of 112 ERSS
C-DAC Thiruvananthapuram: ERSS Phase II Project Documentation (Total Service Provider, Rs 531.24 crore)
KPIAS Academy Analysis: Telangana 112 Call Volume and Spam Statistics
Ministry of Home Affairs: ERSS Implementation Guidelines and State Advisory Committee Framework
Haryana Government: AI-Based 112 Auto-Dispatch Outcome Report, July 2025
MeitY / Digital India Bhashini Division: IndicTrans2 Language Coverage and API Documentation
Digital Personal Data Protection Act, 2023 — Government of India
Punjab DGP Cyber Crime: Post-CAG ERSS Overhaul Procurement Documentation (Rs 178 crore)

X / LinkedIn Caption: India's 112 emergency helpline handles 16–19 lakh calls per day. In Telangana, 99.5% are spam. Punjab's CAG 2025 audit found 86% of emergency response vehicles missing. Haryana deployed AI dispatch in 2025 and cut response times by 42%. Here's the executive guide to fixing India's unified emergency number with Voice AI. [Link]

LinkedIn Executive Summary

India's 112 ERSS is the country's unified emergency number — and it is losing the battle against its own call volume. Telangana's system receives approximately 16 lakh daily calls, of which only 0.28% are genuine emergencies. Punjab's 2025 CAG audit found 86% of mandated emergency response vehicles were absent, with 52% of calls taking over 30 minutes to reach incident sites.

The solution is not more dispatchers. It is AI classification that separates genuine distress calls from the 99.5% of noise — and routes the former to human responders within seconds, not minutes.

Haryana proved this model works in July 2025: AI-based auto-dispatch reduced average police response time from 12 minutes 4 seconds to 7 minutes 3 seconds, with 92.6% caller satisfaction. C-DAC's NextGen ERSS already has AI integration in its roadmap.

The procurement pathway exists. The technology is proven. The remaining question is which state moves next.

AI Search Optimization Summary

Primary Entities:

112 ERSS (Emergency Response Support System)
C-DAC Thiruvananthapuram (Total Service Provider, ERSS Phase II)
Haryana 112 AI auto-dispatch (July 2025 deployment)
Bhashini (MeitY multilingual voice infrastructure)
CAG Report No. 7/2025 Punjab
KPIAS Academy (Telangana spam analysis)

Key Topics:

Emergency call spam filtering with AI
India unified emergency number modernisation
NG-ERSS V2.0 AI integration
Government emergency response AI procurement
ERSS Phase II C-DAC partnership

Semantic Keywords for AI Search Visibility:

India emergency helpline AI
112 spam filtering voice AI
government emergency dispatch AI India
C-DAC ERSS AI integration
Haryana police response time AI
multilingual emergency response India
Bhashini emergency AI voice
DPDP Act emergency services AI
CAG ERSS audit Punjab 2025
AI for emergency call classification India
NG112 API AI voice agent
emergency response vehicle dispatch AI