Startup Deep Dive : Innefu Labs — it claimed 98% accuracy, Delhi Police recorded under 1%

Innefu Labs says its facial recognition software identifies people with 98.3% accuracy. Delhi Police, the agency that has run the software since 2018, told the Delhi High Court its real-world accuracy was under 1%.

That gap sits at the centre of a company that is otherwise having its best year yet. The New Delhi firm builds AI software for defence, intelligence, police and tax agencies, and in June 2026 it closed a $30 million Series B from Singapore-based Panthera Growth Partners — the largest single cheque an Indian “sovereign AI” security vendor has disclosed. Revenue crossed ₹100 crore for the first time in the year before that raise. The company was also bootstrapped for its first seven years, has been named by Amnesty International in a spyware investigation it denies, and was itself breached by a hacker in 2024. All of it is documented, and all of it is below.

Quick facts

What Innefu Labs sells, and to whom

Innefu Labs builds artificial-intelligence software that governments use to identify people, predict crime and trace money. Its buyers are defence and intelligence agencies, state police forces, financial-intelligence and revenue-enforcement units, and a smaller set of large enterprises. The company says it has more than 100 installations across the Indian subcontinent, the Middle East and Southeast Asia, and that its platforms have powered India’s first National Terrorism Data Fusion Centre and what it describes as Southeast Asia’s largest operational intelligence fusion centre (Innefu Labs press release, June 2026; PR Newswire, June 2026). It does not sell to consumers, and it does not publish a retail product list — everything is procured, deployed and serviced project by project, agency by agency.

The origin: an authentication problem at DRDO

Innefu Labs was founded in 2010 by Tarun Wig and Abhishek Sharma, two Delhi-based engineers who between them had spent close to two decades in cybersecurity and enterprise software before starting the company (The420.in, June 2026). Wig had come up through cybersecurity strategy and sales; Sharma through software development and SAP/ERP delivery. The company’s first real product was not facial recognition — it was AuthShield, a two-factor authentication tool that verified logins at the network protocol level rather than inside individual applications, aimed at securing corporate email, SAP systems, VPNs and cloud infrastructure (DQ India, 2015; Innefu Labs press release). The credibility hire came early: the Defence Research and Development Organisation deployed AuthShield to protect its own SAP ERP systems, giving a young private vendor a foothold inside India’s defence-technology establishment (DQ India, 2015; ProQuest trade-journal listing). That authentication and identity-verification base is what the company later extended into biometric recognition and predictive-policing software, the products it is best known for today.

The struggle years

Two things nearly capped Innefu Labs’ growth, at very different points in its life.

The first was capital. The company built and sold AuthShield, and later its analytics products, for seven years on internal revenue alone. Its first institutional funding — a $2 million Series A from IndiaNivesh Venture Capital Fund, with First Bridge Investment Managers also participating — did not arrive until April 2017 (Inc42, 2017; Analytics India Magazine, 2017; Innefu Labs press release, June 2026). For a company selling into government procurement cycles, where payment can trail delivery by months, seven years of self-funded working capital is a long runway to hold without an outside cushion.

The second was reputational, and it centred on the product that made Innefu Labs a household name inside Indian policing. In March 2018, the Delhi High Court, hearing a missing-children case (Sadhan Haldar v. NCT of Delhi), directed Delhi Police to procure facial recognition software; Innefu Labs’ system was the one deployed. By January 2019, the same court was told the software had not helped crack a single missing-child case, and the Ministry of Women and Child Development conceded it could not reliably distinguish a boy from a girl. Delhi Police’s own disclosures put the system’s real-world accuracy at about 2% in 2018, falling below 1% by 2019 — against a claimed accuracy of 98.3% still advertised on Innefu Labs’ own site as of mid-2025 (The Wire, 2 July 2025). The company’s co-founder and CEO Tarun Wig declined to comment when the outlet put the discrepancy to him.

The turning point

For most of its history Innefu Labs was a name known mainly to procurement officers inside security agencies, not to investors. That changed on 5 June 2026, when Panthera Growth Partners, a Singapore-based growth-equity firm investing institutional capital from India, Europe and the United States, closed a $30 million Series B in the company, structured as a mix of primary capital and a secondary share sale (PR Newswire, 5 June 2026; Business Today, 5 June 2026). It was, on the company’s own account, its largest disclosed funding event — roughly fifteen times the size of the $2 million Series A that had taken seven bootstrapped years to arrive.

The numbers on either side of that moment tell the same story from a different angle. In FY24, the year before the raise closed, Innefu Labs reported revenue of ₹62.7 crore and net profit of about ₹18 crore. By FY25 — the last full year before the Series B — revenue had risen to ₹103 crore, up 63.7% year on year, and net profit had risen to ₹34.2 crore, up about 90% (Business Today, 5 June 2026; Inc42 financials, 2026). Panthera was not funding a story; it was funding a company whose government contract pipeline, put at more than ₹100 crore ($10 million-plus) in the same reporting, had already started converting into a very different scale of profit and loss statement. The round also explicitly names IPO preparation as a use of proceeds, alongside international expansion and a new “Physical AI” robotics division (PR Newswire, 5 June 2026; HDFC Sky, June 2026).

The money behind it

• April 2017 — Series A, $2 million: led by IndiaNivesh Venture Capital Fund, with IndiaNivesh Fund Managers, First Bridge Investment Managers and individual investors Divya Rathi and Anju Rajgarhia participating (Inc42, 2017; Analytics India Magazine, 2017).
• June 2026 — Series B, $30 million: led by Panthera Growth Partners (Singapore), a mix of primary investment and secondary stock purchase; the company’s largest disclosed round (PR Newswire, 5 June 2026).
• Total disclosed funding: reported at roughly $33 million across all rounds as of June 2026 — dominated by the Series B (Inc42 company profile, 2026).
• What Panthera changed: Panthera partner Shilpa Kulkarni cited the company’s “proprietary technology, deep domain expertise” and record in “high-stakes, mission-critical environments” as the basis for the check, and the round is explicitly earmarked to fund overseas expansion beyond the company’s existing West Asia presence, sovereign large-language-model development, an agentic-AI push and IPO preparation (PR Newswire, 5 June 2026).
• What First Bridge and IndiaNivesh changed: the 2017 Series A was Innefu Labs’ first institutional validation after seven bootstrapped years, and First Bridge’s Krishan Kant Rathi later said the firm had backed the company for its vision of building AI platforms for India’s “critical sectors” (Innefu Labs press release, June 2026).
• Valuation: not disclosed in either round; databases that track private valuations list the figure as confidential (Tracxn, 2026).

How it makes money

Innefu Labs sells software and the services to deploy it, largely to government buyers, which makes its business look more like a systems-integration contractor than a subscription SaaS company. Revenue arrives through a small number of large, negotiated, multi-year agency contracts rather than many small self-serve accounts — a structure Sacra’s company analysis describes as “high-touch, deployment-heavy” (Sacra, 2026).

• Money in: licensing and deployment fees for platforms such as Prophecy (its big-data-analytics suite, with modules including Alethia for predictive policing and InteleLinx for linking faces to phone and communication records), Innsight for open-source identity cross-referencing, an AI Vision real-time video-analytics platform, biometric (facial and iris) recognition systems, and AuthShield-line authentication products, plus the ongoing maintenance and support contracts that typically follow a government deployment (company product pages, innefu.com, 2026; The Wire, 2 July 2025).
• Costs out: FY25 total expenses were reported at ₹58.2 crore against ₹103 crore of revenue (Inc42 financials, 2026) — a cost base dominated, as with most deployment-heavy government software vendors, by engineering, on-site delivery and compliance staff rather than by paid customer acquisition, since the company does not advertise to consumers.
• Where the margin sits: on the reported FY25 numbers, net profit of ₹34.2 crore on revenue of ₹103 crore implies a net margin of roughly 33% — high for a services-heavy government contractor, and a sign that its long-cycle, low-competition contracts (defence and intelligence work rarely goes out to open tender) carry real pricing power once won.
• What people get wrong: because its most publicised product is facial recognition, it is easy to assume Innefu Labs is primarily a surveillance-camera company. Its own disclosures suggest a wider base — GST and tax-fraud analytics for revenue agencies, financial-intelligence tooling for banks, and cyber-authentication for defence and enterprise IT sit alongside policing work (innefu.com product pages, 2026).

The numbers

Only two years of audited figures are publicly reported with specific rupee amounts; earlier filings sit behind paid corporate-database subscriptions and are not verifiable from public sources, so they are omitted here rather than estimated.

Sources: Business Today, 5 June 2026, citing MCA filings; Inc42 financials, 2026.

Where the money comes from

Innefu Labs does not publish a revenue split by segment or geography, so the picture below is built from what it discloses about its customer base and footprint rather than from percentages, which are not public.

• Government and public safety: defence, intelligence and state police forces, including Delhi Police and Uttar Pradesh Police deployments of its facial-recognition and predictive-policing tools (The Wire, 2 July 2025; Medianama, November 2020).
• Revenue and financial intelligence: tax and GST-fraud detection for revenue-enforcement agencies, plus financial-intelligence-unit and banking clients (innefu.com product pages, 2026).
• Enterprise and defence IT: authentication and cybersecurity products such as AuthShield, with DRDO as a disclosed reference customer (DQ India, 2015).
• Geography: India remains the core market, with the company citing more than 100 installations across the Indian subcontinent, the Middle East and Southeast Asia, and the June 2026 raise earmarked partly to expand beyond its existing West Asia presence (PR Newswire, 5 June 2026).
• The surprise: a company built to reduce India’s reliance on foreign security technology is now scaling on foreign capital — Panthera’s fund draws on institutional investors from India, the European Union and the United States, even as CEO Tarun Wig frames the company’s mission as reducing dependence on “external technologies” to secure India’s institutions (PR Newswire, 5 June 2026).

The risks

• The accuracy gap on its flagship product is documented in court, not just alleged. Delhi Police told the Delhi High Court its facial recognition accuracy was around 2% in 2018 and under 1% by 2019, against Innefu Labs’ own advertised claim of 98.3% as of mid-2025 — a discrepancy the company has not explained publicly (The Wire, 2 July 2025).
• It has been named, and has denied involvement, in a state-linked spyware investigation. In October 2021, Amnesty International reported that infrastructure used to target a human-rights activist in Togo with spyware linked to the “Donot Team” hacking group traced to an IP address associated with Innefu Labs; Innefu Labs denied any link to the spyware or the targeting, and Amnesty said it found no evidence of the company’s direct involvement (Amnesty International Security Lab, 7 October 2021; SecurityWeek, October 2021).
• A cybersecurity vendor was itself breached. In January 2024, a threat actor using the handle “PreciousMadness” advertised roughly 54GB of data allegedly exfiltrated from Innefu Labs’ systems — including access to its Fortinet VPN and Microsoft 365 environment — for sale on a criminal forum, a notable incident for a company whose business is securing other organisations’ data (The Cyber Express, 2024).
• Revenue depends on a small number of government relationships. With no disclosed segment split and a business model built on large agency contracts rather than a broad paying base, the loss, delay or non-renewal of even one or two major contracts would be disproportionately felt (Sacra company analysis, 2026).

The takeaway

The lesson in Innefu Labs’ story is not that government-facing AI companies are fragile — its financials say the opposite, with profit growing faster than revenue in the year before its largest raise. It is that being unaccountable to a retail market cuts both ways. A consumer-facing company whose flagship product performed at under 1% of its advertised accuracy would likely not survive the public reaction. A government vendor can keep advertising the higher number for years, because its customers are agencies, not the people the product is used on, and the people it is used on rarely have standing to demand an explanation. The commercial success is real and documented. So is the accountability gap that let one number sit unchallenged next to another for the better part of a decade.

Frequently asked questions

What does Innefu Labs do?

It builds AI software for governments — facial and iris recognition, predictive policing, GST and financial-fraud analytics, and cyber-authentication tools — sold mainly to defence, intelligence, police and revenue agencies in India and, increasingly, the Middle East and Southeast Asia (PR Newswire, 5 June 2026).

Who funds Innefu Labs?

Its disclosed backers are IndiaNivesh Venture Capital Fund and First Bridge Investment Managers, who backed its 2017 Series A, and Panthera Growth Partners, which led a $30 million Series B in June 2026 (Inc42, 2017; PR Newswire, 5 June 2026).

Is Innefu Labs profitable?

Yes, on its most recent disclosed numbers. It reported net profit of ₹34.2 crore on revenue of ₹103 crore in FY25, up from about ₹18 crore of profit on ₹62.7 crore of revenue in FY24 (Business Today, 5 June 2026).

Is Innefu Labs a listed company?

No. It is privately held. Its June 2026 funding round explicitly names IPO preparation as one intended use of proceeds, but no listing date or exchange has been announced (PR Newswire, 5 June 2026).

What is the controversy around Innefu Labs’ facial recognition technology?

Delhi Police, which has used the company’s software since 2018, disclosed to the Delhi High Court that its real-world accuracy was around 2% in 2018 and under 1% by 2019 — far below the 98.3% accuracy Innefu Labs advertised as of mid-2025. Separately, in 2021 Amnesty International linked infrastructure traced to the company to a spyware campaign against a human-rights activist, a link Innefu Labs has denied (The Wire, 2 July 2025; Amnesty International Security Lab, 7 October 2021).

Sources

Figures are as of September 2026. Currency converted at $1 ≈ ₹96.0 as of 18 September 2026 (Trading Economics).

• Business Today, “National security AI startup Innefu Labs bags $30 million funding”, 5 June 2026
• PR Newswire (Panthera Growth Partners), “Panthera Growth Partners Invests USD 30 Million in Innefu Labs Series B Round”, 5 June 2026
• Innefu Labs press release, “Innefu Labs Raises USD 30M in Series B Round to advance Sovereign AI for National Security”, June 2026
• The420.in, “Tarun Wig-Led Innefu Labs Secures $30 Million to Expand National Security AI Push”, June 2026
• Inc42, company financials page for Innefu Labs, accessed 2026
• Inc42, “Artificial Intelligence Startup Innefu Labs Raises $2 Mn Series A From IndiaNivesh”, 2017
• Analytics India Magazine, “Delhi AI startup Innefu Labs secures $2 million from IndiaNivesh Fund”, 2017
• HDFC Sky, “AI Firm Innefu Labs Bags USD 30 Million In Funding from Panthera Growth Partners; Plans IPO to Expand Globally”, June 2026
• The Wire, “As AI Took Over Policing in Delhi, Who Bore the Brunt?”, 2 July 2025
• Medianama, “Exclusive: Delhi Police working on arming PCR vans with facial recognition software”, November 2020
• Amnesty International Security Lab, “Togo: Prominent activist targeted with Indian-made spyware linked to notorious hacker group”, 7 October 2021
• SecurityWeek, “Amnesty Links Indian Cybersecurity Firm to Spyware Attack on African Activist”, October 2021
• The Cyber Express, “Innefu Labs Data Breach: Major Cyberattack Hits Indian Cybersecurity Firm”, 2024
• DQ India, “DRDO implements Innefu AuthShield Two Factor Authentication”, 2015
• Sacra, company analysis of Innefu Labs, 2026
• Tracxn, Innefu Labs company profile, 2026
• Innefu Labs company product pages, innefu.com, 2026

Found an error? Write to us and we’ll correct it in the open, dated, on the piece.