The UK Government’s One Login initiative perfectly illustrates how centralised identity systems fail.
Despite reaching 13 million users with fanfare, the system lost its trust framework certification in early 2025 as key suppliers walked away. Security assessments revealed data protection issues, and the system still hasn’t fully implemented the government’s own Secure by Design principles — with reports of poor oversight and major data protection failings. The pattern is a symptom of a common design flaw.
This is a sadly familiar story. A well-intentioned effort to simplify public infrastructure runs aground and begins to lose public trust. In many cases, like Horizon IT consolidation, the loss of trust is permanent and leads the simplification effort to fail. One Login could easily become the latest iteration of that mistake. But there’s an opportunity to rewrite the ending of this tale.
The first step is to acknowledge the core design problem: centralisation. This is especially true for identity systems, which involve sensitive personal and sometimes biometric data.
For many years, centralised technology architectures were the default choice for governments and large corporations. One might have argued that the failures of trust were an unfortunate but unavoidable consequence of technology.
But the technology has changed, and so must the choices of public authorities. The future of digital identity — and the One Login program — must be decentralised, standardised, and user-controlled by design.
Centralisation is Fragility by Design
Digital identity, when managed through a centralised authority, becomes a single point of failure technically, politically, and ethically. If one account is compromised, or one developer access misused, the consequences can quickly ripple across millions of users. More importantly, individuals are left with no meaningful recourse or control when something goes wrong.
The failure of Horizon, the Post Office’s accounting platform, wasn’t just about bad code. It was about a system built without checks, transparency, or user agency. One Login risks the same fate. When platforms, including government platforms, hold disproportionate control over access to essential services, the public ends up navigating opaque, brittle systems that collapse under scrutiny.
And that is just the tip of the iceberg. Data integrity errors — bad or contradictory records — do far more silent damage. Most UK tax dispute settlements reviewed by HM Revenue & Customs (HMRC) in 2021 fell below its own quality standards, with 10% causing financial harm to taxpayers. Meanwhile, over a third of the telephone calls made to HMRC are progress chasing or calling to correct HMRC errors, suggesting endemic data quality problems across government systems.
When every service draws from the same flawed source, errors don’t just persist — they multiply and become increasingly expensive to fix. A wrong address becomes gospel truth across dozens of departments with no cost-effective way to clean it up, the definition of an integrity breach. Centralisation is fragility by design.
Identity is a Relationship, Not a Record
One of the biggest misconceptions in digital infrastructure is treating identity as a static credential — something to be stored, retrieved, and verified like a barcode. But identity is fluid. Who I am to my healthcare provider is not who I am to my bank, or my local council. Digital identity is a set of contextual interactions. Therefore it must be architected as a relationship, not a record.
Centralised systems like One Login try to flatten those nuances. These systems consolidate credentials into a single format, managed by a single gatekeeper. However, the single gatekeeper model doesn’t reflect how humans operate and it doesn’t scale ethically or securely.
If this sounds abstract, consider today’s consumer experience with digital wallets. People use mobile wallets for payments, tickets, even digital IDs. But these wallets are delivered by massive corporations that centralise the data of millions of customers around themselves, instead of allowing the customer to manage and organize their own information in the way that’s best for them.
For the customers, that means airline tickets live in one app. Vaccine records in another. Loyalty cards, keys, insurance are all scattered across proprietary silos. The result? Individuals are left juggling a dozen disconnected tools, logins, and data that doesn’t talk to each other or provide any cross benefits. This is exactly the kind of ecosystem centralised platforms create: fractured, inefficient, and extractive. The paradox is that centralized platforms create fragmentation at the user level by forcing citizens to navigate between silos that don’t communicate.
From Digital Wallets to Agentic Infrastructure
A different approach is needed, one that treats personal data not as a resource to be aggregated, but as something individuals should control directly. This means shifting away from platform-owned identity systems toward models where data is stored in environments individuals manage themselves, and access is granted selectively, based on purpose and context.
Ironically for the One Login project, other parts of the Department for Science, Innovation and Technology are already embracing the decentralised approach to identity. The Digital Identity and Attributes Trust Framework (DIATF) was first launched in 2023 to establish rules that any business could follow in order to be deemed a trustworthy verifier of identities. The rules emphasize principles that empower consumers like privacy, transparency and interoperability. DSIT is about to release the fourth version of this trust framework, and has certified more than 50 companies against it. Notably, One Login itself lost certification earlier this year when key suppliers withdrew from the framework. DSIT Secretary Peter Kyle recently announced that the government’s own GOV.UK Wallet will feed into the decentralised DIATF ecosystem, allowing citizens to put their government-issued ID cards to use in the private sector in the ways they choose.
DSIT also plans to integrate its generative AI chatbot into the GOV.UK Wallet. This combination of decentralised identity with AI systems — while still cautious and rudimentary in the government’s plan — is an important step toward avoiding the known pitfalls of centralisation in the era of AI.
As AI systems take on more decision-making roles, the ability to delegate AI agents to act in our name, with appropriate guardrails, becomes essential. The underlying architecture of AI “identity” must support transparency, revocability, and separation of concerns. AI agents must be able to bring personal information into their interactions safely, without needing to surrender it to centralised gatekeepers in order to function. In this model, intelligence doesn’t require centralisation; it requires structure that respects autonomy and enables alignment.
Identity at the Heart of AI
AI systems are beginning to operate with greater independence. They are reasoning, summarising, and increasingly acting on our behalf. But intelligence alone is not enough. For these systems to be genuinely useful, they must operate with context: a clear understanding of an individual’s preferences, constraints, goals, and history. That level of contextual awareness is difficult to achieve within today’s fragmented application landscape or centralised identity systems that abstract users into static profiles.
What’s needed is a model where individuals can selectively share relevant information spanning public, private, and transactional domains without giving up long-term control. This enables AI systems to make decisions or take actions that reflect the person’s actual circumstances, while preserving boundaries. Critically, such a model must allow information to be accessed without being copied or absorbed into another platform aligning usefulness with accountability, and intelligence with consent. This is what it means to treat identity as a relationship, not a record, in the AI age.
The Way Forward
Efforts like One Login may begin with the right intentions, but without changes to their underlying architecture, they will continue to lose public trust and fail to deliver on the very real benefits they promise. The choice of centralised technology always leads to the pitfalls of centralised control: poor oversight, lack of accountability and unintentional harm to citizens. A more sustainable path is both possible and increasingly necessary. It’s one that restores control to individuals, enables meaningful interoperability across systems, and supports intelligent tools without compromising autonomy.
The technical foundations for such a shift already exist. What’s required now is a commitment to building systems that prioritise structural transparency, verifiable consent, and resilience over time. Digital identity should not be reduced to a single login. It should function as a personal infrastructure and be responsive to context, aligned with user intent, and capable of supporting a wide range of interactions without sacrificing personal control.
The challenge is no longer conceptual. It’s political and institutional. And it demands leadership willing to treat trust not as a message, but as an outcome of design. The future of digital identity is not a single login. It’s a personal infrastructure owned by you, working for you, and serving only you.
Let’s build it.
About the Author
Davi Ottenheimer is VP of Trust and Digital Ethics at Inrupt. Sir Tim Berners-Lee, inventor of the World Wide Web, created Solid to realize the web as he fully envisioned it. Sir Tim co-founded Inrupt to provide enterprise-grade Solid software and services. Inrupt’s data infrastructure software enables enterprises and governments to deploy and manage Solid-compliant solutions. Our products are the expression of decades of experience in security, compliance, and operational excellence.
