Image credit: Exceptional Technology — AI-generated illustrative scene for Defmate Secure, a software platform with no official product photograph available.
Every day, in offices, ministries and headquarters across India, someone pastes a report, an email or a plan into an AI assistant to work faster. It is a small, ordinary moment — and, the company says, it is where the greatest security risk now lives. On 3 November 2026, Gurugram-based Defmate unveiled a platform built for exactly that moment: Defmate Secure, an AI security layer that checks every message on the user’s own device, before it ever leaves.
The idea is beautifully simple. Most AI security tools begin their work after information has already gone out — encryption, retention limits, data policies. Defmate Secure starts earlier, with a single question: should this information leave the organisation at all? If the answer is no, the sensitive details are masked or held back, the user sees a short reason, and is offered a safer version to send. Nothing is recalled after the fact, because nothing sensitive ever left.
Privacy before disclosure
Defmate Secure evaluates each message before it is sent to any AI model, public or private, the company says. Each organisation defines what counts as sensitive for its own people, so protection follows its own security policy rather than a generic rulebook. The platform is designed to complement encryption, access controls and existing cybersecurity — adding a layer that governs what is shared in the first place.
That matters because the most dangerous leaks are rarely a single secret typed in plain sight. In defence, the company points out, an officer’s rank is rarely secret, a unit’s name may be public, and a location may already be in the news — but combined with a date, they can describe an operation. Intelligence professionals call this the mosaic effect, and Defmate Secure is designed to recognise sensitive combinations as well as individual details, including fragments that appear across several messages in one conversation.
More Read
The same logic, the company says, applies far beyond defence. A client’s name, a deal value and a closing date can move a market. A patient’s name and a test result can break a duty of confidentiality. Defmate Secure is built for every institution that handles sensitive information: government departments, banks and financial institutions, hospitals, critical infrastructure operators, and enterprises guarding confidential strategy, client data or intellectual property.
Born from defence, built for everyone
Defmate Secure has its origins in Defmate’s work on sovereign AI and decision support for the Indian Armed Forces, where the company says confidentiality, reliability and human judgement are non-negotiable. Those requirements shape every system the company builds — and they explain why the platform is designed the way it is.
Defmate is a defence technology company based in Gurugram, founded by alumni and professors of the Indian Institutes of Technology. Its core engine, Nirnay, is a proprietary rule-based decision support system designed and built in-house: a prediction from any model enters Nirnay as an input, never as an instruction, and a separate, fully deterministic system is the only thing that produces a verdict, against a rule set the deploying authority approves and a named officer signs. Every decision is sealed into a tamper-evident audit chain, the company says.
Around that core, Defmate builds a family of small language models, each sized to a single job — from a 4-billion-parameter model for a forward node with no signal, through 12- and 31-billion-parameter tiers, up to a 70-billion-parameter tier for a command post — with the whole stack designed to run fully offline for weeks if required. The company applies four tests to the word sovereign, and holds that all four must pass at once: the data stays inside Indian systems; the system runs on Indian infrastructure, including with no connection at all; India owns the code and the rules; and a named human can see the reasoning and overrule it.
Use AI, keep control
For organisations that need sensitive work to remain entirely in-house, Defmate Secure pairs with Defmate’s sovereign AI, which runs on the organisation’s own network and can operate fully offline. The company’s aim, it says, is not to keep institutions away from AI but to let them use it with confidence. “Organisations should not have to choose between capable AI and control of their own information,” a Defmate spokesperson said. “Defmate Secure returns that choice to where it belongs: the institution, and the people accountable for its decisions.”
As AI moves deeper into institutional work, the company argues, control over what organisations disclose will matter as much as protecting what they store. Defmate Secure is the first step in its broader programme to build sovereign AI and decision-intelligence capabilities for mission-critical organisations — Indian technology, built for India’s own requirements, that lets powerful AI in without letting secrets out.
Keep reading: Eyes That Never Blink: Inside Oculosense’s Drishti Vision Live AI Glasses · Seeing Through Smoke and Darkness: Inside Nova Spectral AI’s Quantum-Dot Chip That Gives Machines Night Vision · More Industrial Automation & Technology stories
Like stories like this? Add Exceptional Technology as a preferred source on Google to see more of our reporting in your search results and Discover feed.
