Digital Identity: Bangladesh Bets $748 Million on Trust

Here's the part that surprised me: Bangladesh is spending $748 million to build a national identity system, and the expensive part isn't the face scans. It isn't the fingerprints either. Eighty percent of that budget is going toward printing 200 million chip cards and collecting biometric data (your face, voice, fingerprints, basically the parts of your body that make you uniquely you). That's the visible, photogenic part of the project. But according to Biometric Update, the part that actually determines whether this thing works, or becomes a $748 million paperweight, is something almost nobody talks about: getting a tax office, a hospital, a bank, and a government service to all agree on "yes, that's the same person" without each one collecting a copy of everything about them. That's what a digital identity really is. Not a card. Not a face scan. A trust system.

What digital identity actually means, and why Bangladesh is paying so much for it

Let's start with what most people assume, because it's a totally reasonable assumption. You hear "national digital identity project," and your brain pictures one giant government database with everyone's face, fingerprints, address, and medical history sitting in a folder somewhere. That's the misconception, and it's an easy one to fall into. News coverage of these projects tends to focus on enrollment day: the cameras, the fingerprint scanners, the new plastic card. That's the dramatic part. It photographs well. Nobody puts "backend authentication protocol" on the evening news.

But here's the reality. Bangladesh's One-ID project is explicitly modeled on pieces of Estonia, Singapore, and India's identity systems, and what those countries figured out the hard way is that the card is basically a receipt. The real engineering challenge is interoperability, meaning making totally separate computer systems (a bank's system, a hospital's system, a tax office's system) recognize and trust the same identity without each one needing its own full copy of your file. Bangladesh's plan links tax identifiers and health identifiers into one profile, then connects that profile to a mobile wallet for payments. Every one of those connections is its own technical negotiation. That's where the money and the years go.

Digital identity interoperability explained in one sentence

Interoperability just means different computer systems can check the same identity fact and get the same trustworthy answer, without passing your entire file back and forth every time someone needs to confirm something small.

---

How does a national digital identity system actually verify a person without oversharing?

Short answer: it splits the process into three separate jobs, and only one of them ever touches your full record. Step one, an authority (in Bangladesh's case, the government) confirms you're a real, unique person and issues you a credential. Step two, when a service needs to check something about you, it doesn't ask for your file, it asks a specific yes-or-no question. Step three, you get a confirmation back, not a data dump.

This is the piece that clicked for me once I understood it, and it's called selective disclosure. It's a way of proving one specific fact (say, "this person is over 21") without handing over the birthdate, the address, the ID number, or anything else sitting on the credential. According to research on identity architecture from walt.id, this isn't a nice-to-have feature bolted on for privacy points. It's a structural requirement. Updated European identity regulations actually require selective disclosure specifically so that data minimization (collecting only what's needed, nothing more) is built into the system, not left to good intentions.

Think about the last time a bartender checked your ID. If they're doing it right, they're checking exactly one fact: are you over 21? But your physical driver's license doesn't know how to answer just that question. It hands over everything at once: your full name, your home address, your exact birthdate, your license number. The bartender didn't need any of that. They needed a yes or a no. Physical ID cards overshare because they're a dumb piece of plastic. A well-built digital identity system is designed to do the opposite: answer only the question that was asked. This article is part of a series, start with Uk Age Verification 1 400 Vpn Privacy Signup Surge Podcast.

---

Bangladesh one-id project shows why consent-based data sharing is an engineering problem, not just a policy

Here's where it gets interesting. Bangladesh's One-ID is built to run on what's called consent-based data sharing, backed by the country's Personal Data Protection Act and its National Data Governance policy. That sounds like a legal detail. It's actually a software requirement disguised as a legal one.

Think about what "consent-based" has to mean in practice. It can't just be a checkbox you tick once when you sign up. It has to be a decision the system makes every single time an office asks for something, because a hospital asking to confirm you're insured today is a completely different request than a bank asking to confirm your income last year. According to the researcher analysis drawing on Biometric Update's coverage of Bangladesh's broader digital trust strategy, this means every transaction needs a live permission flow built into the plumbing, not a policy sitting in a filing cabinet somewhere saying "we promise to ask nicely."

This is also, weirdly, where digital identity systems from birth get complicated in a good way. Bangladesh plans to automatically generate a digital ID for a newborn, linked to the parents' national ID. That sounds simple on paper. But it means the system has to track a chain of trust across decades, updating who can access what as a child becomes an adult, gets a job, opens a bank account, and eventually has kids of their own. Every one of those life stages changes what should and shouldn't be shared. That's not a database problem. That's a lifetime of access-control decisions, encoded in software.

Old-style identity check	Interoperable digital identity system
Hands over full ID document	Confirms one specific fact only
Each office stores its own copy	No permanent copy stored by the verifier
One breach exposes everything	Breach exposes only the fact requested, not the file
Consent given once, applied everywhere	Consent checked at every transaction

Digital identity theft risk versus selective disclosure

The risk with old-style ID checks is that every organization ends up storing a duplicate of your personal file, and every duplicate is a new target for a breach. Selective disclosure lowers that risk by making sure most verifiers never receive or store the underlying data at all, just a cryptographic proof (a scrambled, verifiable answer) confirming one fact.

---

The misconception about digital identity that even smart people fall for

Let's name the misconception directly, because I think it's genuinely reasonable to have believed it. Most people assume a digital identity system is basically a giant, centralized list: everyone's face, fingerprints, and personal details sitting in one place, accessible to whoever has the password. It's an easy thing to believe because that's often how these projects get described in headlines, and because enrollment (the scanning, the photographing, the fingerprinting) is the only part regular people ever see or experience directly.

But that mental picture misses the actual architecture. A well-designed digital identity system minimizes how much any single party holds. The government authority that issues your credential holds the full record, yes, because someone has to be the source of truth. But every other party in the system, your bank, your clinic, your employer, your streaming service, is designed to receive as little as possible. Bangladesh's plan explicitly frames this as building "trust infrastructure," not a data warehouse, because that determines whether a verifier receives an entire record or a limited proof.

Fragmentation is the baseline problem interoperability solves. Without it, people repeat the process of obtaining and verifying their personal information across every single network they touch. Previously in this series: Digital Identity Verification Proving 18 Without A Birthday .

summarized from research on national identity architecture, arXiv

Why does that framing matter to you, sitting on your phone, not building any of this software? Because it means the question you should be asking about any identity system, whether it's a national ID or an app asking you to verify your age, isn't "how much of my data do they have." It's "how much of my data does the thing checking my identity actually need to see." Those are very different questions, and only one tells you what the verifier receives and can store.

This is a distinction I think about constantly in the facial recognition work we do at CaraComp, because it's the same principle at a smaller scale. A face scan that confirms "this is the account holder" and immediately discards the image is a completely different privacy situation than a face scan that gets stored, tagged, and cross-referenced forever. The technology looks identical from the outside. The architecture underneath it is what actually determines your risk.

Bangladesh one-id project and the age-verification comparison

The same selective disclosure principle that powers Bangladesh's One-ID is what a well-built age check should use online, confirming you're over a certain age without storing or transmitting your exact birthdate, name, or ID number to the site doing the asking.

---

Why the aha moment about digital identity actually protects you

So here's the aha moment, the thing I want you to walk away actually able to explain to someone else. When a system is engineered so that different offices can confirm your identity without each collecting a duplicate of your file, the convenience and the privacy protection aren't two separate features. They're the exact same piece of engineering. The thing that makes life easier (not re-entering your information at every office) is the thing that also makes you safer (nobody's stockpiling a copy of everything about you).

That's counterintuitive, honestly. We're trained to think convenience and privacy trade off against each other, that you sacrifice one for the other. Digital identity, done properly, breaks that trade-off. The interoperability that lets a clinic and a bank both recognize "yes, same person" is built using the same cryptographic proofs that keep the clinic from ever seeing your bank balance and the bank from ever seeing your medical history.

Next time someone tells you a country is spending three-quarters of a billion dollars on "biometric ID cards," you'll know to ask the better question: what happens after the card gets scanned? Does the system hand over your whole file, or does it just whisper "yes, that's them" and walk away? That whisper, not the card, is the $748 million part. Up next: Credit Card Age Verification Steam Skips Age Estimation.

digital identity: Frequently Asked Questions