Who Is Acting in Our Name? Agentic AI, Deepfakes, and the Case for Sovereign Digital Identity.

Feature Article Who Is Acting in Our Name? Agentic AI, Deepfakes, and the Case for Sovereign Digital Identity.

MON, 02 NOV 2026

There is a quiet assumption buried inside almost every digital system we use: that the thing on the other end is either a person or a script. A human logs in. A program runs. For thirty years, that binary held well enough. In 2026 it is breaking, and it is breaking from two directions at once.

From one side come agentic AI systems: software that does not merely answer questions but books, pays, deploys code, reads inboxes, and calls other software, often with credentials borrowed from a human or a company. From the other side come deepfakes: synthetic faces, voices, and documents convincing enough to pass for the real thing in a bank onboarding flow or a video call with the finance director.

These look like two problems, one for AI safety teams and one for fraud teams. I want to argue that they are one problem wearing two costumes. In both, we have lost the ability to answer a simple question:

Who, or what, is really acting here, and on whose authority?

The most serious answer available is not a better chatbot filter. It is infrastructure: verifiable, citizen-controlled, nationally governed digital identity that extends to machines and to media. That is what I mean by sovereign digital identity, and I believe researchers at AI companies, regulators in Accra and Brussels, and builders everywhere should treat it as safety work, not paperwork.

Part One: The Agent Is an Identity, and We Are Treating It Like a Ghost

The pace of adoption explains the urgency.
Gartner has projected that 40 percent of enterprise applications will embed task-specific AI agents by the end of 2026, up from under 5 percent in 2025. Security practitioners keep repeating a point that sounds technical but is deeply political: every agent is an identity. To do its job it needs keys, tokens, and permissions, and the more tasks we hand it, the more access it accumulates.

What worries me is the gap between that adoption curve and the governance behind it. A 2026 industry survey from Gravitee reported that only about 14 percent of organisations say their agents go live with full security approval. That means most agents in the wild were switched on by a team in a hurry, often under shared API keys, with no clear owner and no clean way to switch them off.

Standards bodies have noticed. In February 2026, NIST's Center for AI Standards and Innovation launched its AI Agent Standards Initiative, and the National Cybersecurity Center of Excellence released a concept paper on agent identity and authorization, proposing that agents be treated as identifiable entities within enterprise identity systems rather than anonymous automation.

The threat catalogue is now documented rather than speculative. The OWASP Top 10 for Agentic Applications, published in December 2025 with input from more than a hundred researchers, codified risks that traditional model-safety thinking underweights.

Three are worth naming plainly:

Goal hijacking through prompt injection. An attacker hides instructions in a web page, a support ticket, or an email. The agent reads it as part of its job and obeys. The attacker never needed a password; they needed the agent to read.
Excessive agency and overprivilege. Agents inherit the structural flaw of service accounts: too much access, held too long, with too little audit trail. One 2026 enterprise survey attributed 37 percent of non-human identity incidents to overprivileged identities.
Tool-chain and protocol weakness. The Model Context Protocol and agent-to-agent frameworks are becoming the plumbing of the agent economy, and security researchers, including the NSA's AI Security Center, have flagged real gaps in how they are deployed.

The deeper shift is conceptual.
Traditional AI risk asked what a model might say.

Agentic risk asks what a system will do, at machine speed, across every system it can reach, without waiting for a human to click "approve" at each step.

A hallucinated paragraph is an embarrassment.

A hallucinated wire transfer is a loss.
So the safety question cannot stop at alignment inside the model. It must include accountability around the model:

A unique, cryptographically verifiable identity for each agent.
Permissions scoped to a single task and expiring quickly.
A named human or legal entity answerable for what the agent does.
A revocation button that actually works everywhere the agent's credentials were accepted.

Visa's Trusted Agent Protocol, introduced in October 2025 for agent-initiated payments, is an early example of the right instinct: prove that the agent is authorised before money moves.

Part Two: The Face on the Screen Is No Longer Evidence

Now turn to the other half of the problem.
Deepfake fraud has moved from curiosity to line item. Estimates differ by methodology, which is itself worth saying honestly: one analysis put US deepfake-related fraud losses at about 1.1 billion dollars in 2025, tripling from 2024; another global tally, drawing on incident databases, counted roughly 2.5 billion dollars in 2025 and a further 764 million dollars in the first half of 2026 alone.

The FBI's 2025 Internet Crime Report, for the first time, carried a dedicated AI-related category: more than 22,000 complaints and roughly 893 million dollars in losses.

Signicat's data showed deepfakes growing from about 0.1 percent of fraud attempts three years ago to around 6.5 percent.

The Arup case, where an employee was tricked by a video call full of synthetic colleagues into moving about 25.6 million dollars, remains the cautionary tale every finance team now knows.

We should not treat any single number as gospel. But the direction is not in dispute, and neither is the lesson:

Seeing and hearing are no longer proof.

Detection tools matter, yet detection is an arms race in which generators improve at least as fast as detectors. A defence that depends on spotting flaws will eventually lose to a flawless fake.

This is why the serious conversation has shifted from "is this fake?" to "can this prove it is real?"

That is the logic of content provenance, and specifically the C2PA standard behind Content Credentials: media is cryptographically signed at the point of capture or creation, edits are logged, and anyone can verify the chain.

The Content Authenticity Initiative reports more than 6,000 members, including camera makers and the largest AI and platform companies. Regulators are converging on the same expectation. India's IT Rules amendment, effective 20 February 2026, requires provenance metadata on synthetic content. The EU AI Act's transparency obligations for AI-generated content phase in from August 2026. Gartner listed digital provenance among its top strategic technology trends for 2026.

I want to be honest about provenance's limits, because overselling it is how good infrastructure loses public trust. A signature proves origin and edit history, not truth. A genuine camera can photograph a staged scene. A deepfake pasted into a screenshot and forwarded on WhatsApp carries no credentials at all, which is exactly how most of us in Ghana and across the diaspora actually consume media.

The C2PA community's answer, durable credentials that pair signed metadata with watermarking and fingerprinting, is promising but unfinished. And live attacks, such as a face-swap during a video call, cannot be solved by labelling files. They require authenticating the session and the person.

Part Three: Why “Sovereign” Is the Operative Word

Here is where the two threads meet.
Verifying a human in a live interaction, verifying an agent acting for that human, and verifying the media they produce all depend on the same foundation: a trustworthy root of identity that the citizen controls and the nation governs.

By sovereign digital identity I mean three things.

First, Citizen-Controlled

The person holds their credentials in a wallet and discloses only what a transaction needs, proving "over 18" or "this is my account" without handing over a photocopy of everything.

Second, Nationally Governed

Identity is issued and regulated under domestic law, with an accountable public institution rather than a foreign platform deciding who counts as real.

Third, Interoperable

Digital identity must be trusted across borders because our economies and our diasporas do not stop at the frontier.

Europe is building exactly this. Under eIDAS 2.0, every EU member state must make at least one European Digital Identity Wallet available to citizens by the end of 2026, with acceptance duties for large platforms and regulated sectors following in 2027.

Whatever one thinks of Brussels, the design principle is sound:

Credentials in the user's hands, issued by accountable authorities, verifiable by anyone entitled to check.

Ghana's Opportunity and Exposure

Ghana is closer to this moment than people realise, and also more exposed.

The Ghana Card is the backbone of banking, mobile money, SIM registration, and public services. That makes it enormously valuable and an equally attractive target.

A widely shared LinkedIn warning from a Ghanaian fraud-prevention practitioner described how the static photo on a card can now be animated to pass a remote liveness check.

A Modern Ghana commentary citing the Bank of Ghana's 2024 fraud report noted identity-theft losses rising from GHS 0.6 million to GHS 5.7 million in one year, which the report reportedly linked to weak due diligence in Ghana Card-based transactions. I would treat that figure as indicative rather than final, but the diagnosis rings true.

Another Modern Ghana analysis of the SIM registration exercise made the sharper point: you cannot use AI to clean up a register that was never verified against the authoritative source.

The lesson is that verification at the moment of registration beats forensics afterwards.

To its credit, the National Identification Authority has moved. It made biometric verification compulsory for Ghana Card transactions in July 2026 and, as reported today, has begun enforcing the National Identity Register (Amendment) Regulations, 2026, which prohibit institutions from photocopying or scanning the card and warn of fines.

That is the right direction: verify against the register through the API, instead of scattering copies of a national identity across filing cabinets and WhatsApp groups.

The question for the next phase is whether that biometric check is built to resist the new generation of attacks, namely injected video, real-time face-swaps, and synthetic documents.

Part Four: A Practical Agenda

I am not interested in a manifesto that ends with "stakeholders must collaborate."

Here is what I would actually build, ordered roughly from most achievable to most ambitious.

1. Register Agents the Way We Register Companies

Every agent that transacts in a regulated space, such as banking, mobile money, procurement, or health records, should carry a verifiable credential binding it to an accountable human or legal entity, a defined scope, and an expiry.

Existing standards like OAuth 2.1, OpenID Connect, and SPIFFE-style workload identity already point this way; the missing piece is national and sectoral policy that makes it normal.

2. Make High-Stakes Actions Require a Human Signature

Moving money above a threshold, changing account recovery details, or altering a land record should demand a fresh, strongly authenticated approval from the human principal, not the agent's say-so.

Agents can prepare; people must authorise.

3. Harden Liveness Against Injection, Not Just Imitation

Biometric checks must assume the camera feed itself can be forged.

That means device attestation, hardware-backed capture, challenge-response designs, plus continuous monitoring rather than a one-time selfie.

4. Sign Official Communications

The Electoral Commission, the Bank of Ghana, ministries, major newsrooms, and telecom operators should publish cryptographically verifiable statements so that a fake directive or a synthetic voice of an official can be checked in seconds.

A public verification page that every citizen knows is worth more than a thousand awareness posters.

5. Use Wallets and Selective Disclosure, and End the Photocopy Culture

The NIA's ban on copying the Ghana Card is a cultural reset as much as a legal one.

The logical next step is a citizen-held credential wallet where a bank learns what it needs and nothing more.

6. Build Regional Interoperability Early

ECOWAS and the AfCFTA framework are natural homes for mutual recognition of digital credentials.

Doing it after fragmented national systems harden will be far more expensive.

7. Fund the Boring Parts

Revocation infrastructure, audit logs, incident reporting, and public verification tools are not glamorous, but they are where trust is actually won.

Part Five: The Objections Deserve Straight Answers

“This Is Surveillance With Extra Steps”

It can be, if built carelessly.
A central register that answers every query about every citizen is a honeypot and a tracking tool.

The sovereign model I am describing leans the other way: selective disclosure, minimal data, strong audit trails, independent oversight, and enforceable penalties for misuse under data protection law.

The NIA's own ban on photocopying is an example of reducing exposure, not increasing it.

“It Will Lock Out the People Who Most Need Services”

This is the objection I take most seriously.
Ghanaians who struggled to obtain or correct a Ghana Card know that identity systems can exclude the poor, the rural, the elderly, and those with mismatched records.

Any verification regime must include human fallback channels, fast correction processes, and a rule that no one is denied essential services solely because a machine check failed.

“Provenance Doesn't Prove Truth”

Correct, and I said so above.
Provenance narrows the problem from "is any of this real?" to "who is vouching for it?"

Truth still requires journalism, courts, and judgement.

But a world where authentic material can be proven authentic is a world where lies lose their cover.

“Won't the Criminals Just Adapt?”

Of course.
Security is never finished.
The point is to shift the economics: when forging an agent's authority or a ministry's statement requires breaking cryptography rather than editing a screenshot, most fraud stops being cheap.

A Note to the Researchers

If you work at an AI lab, here is the invitation.

The most valuable safety research for the next two years may not be a new training technique. It may be answers to questions like these:

How should an agent prove it is acting within a human's delegated authority, and how does that proof survive a chain of five agents calling each other?
How can a system distinguish a user's genuine instruction from instructions injected through content the agent was merely asked to read?
What does a revocation signal need to look like so that it propagates to every service in seconds, not days?
How do we design proof-of-personhood and liveness that works for people with low-end phones, poor connectivity, and no credit history, which describes a very large share of humanity?
How do provenance systems survive screenshots, re-uploads, and messaging-app compression?

These are not Silicon Valley problems.
They are problems for a market trader in Kumasi paying through mobile money, a nurse in Accra whose voice could be cloned, and a diaspora engineer in Manchester wiring funds home.

Systems designed with only the first world in mind will fail where the stakes of fraud are highest and the safety nets thinnest.

Conclusion: Trust Is Infrastructure

We built roads before we trusted trade, and registries before we trusted land.

Digital trust is no different.
In the agentic era, an economy in which software acts on our behalf and any face can be forged will not run on vibes and good intentions. It will run on verifiable identity: for people, for the agents who serve them, and for the media that carries their words.

The nations that treat this as core infrastructure, built with privacy, inclusion, and sovereignty at the centre, will be the ones whose citizens can safely use the extraordinary tools now arriving.

The ones that treat it as an afterthought will be paying for the fraud, in cedis and in confidence.

Ghana has a Ghana Card, a regulator now enforcing biometric checks, and a young, technically ambitious population. That is a real foundation.

The task now is to build on it before the fakes outrun the facts.

Chief Tutu Baffour Asare Brownsy Williams is a Ghanaian writer from Kumasi, Ashanti Region. Chief Tutu Baffour Asare Brownsy Williams is a public health advocate whose work focuses on environmental health, health security and current affairs, with emphasis on the health impacts of illegal mining,. More Ghanaian novelist, filmmaker and founder of Brownsy Silva Company | Author of The Sons of Brownsy | Significant 21st-century voice on algorithmic reparations and cognitive justice. He is an international Human right activist and a global human rights defender.

*Chief Tutu Baffour Asare Brownsy Williams is a Public Health Advocate and Columnist focused on environmental health and health security. His work tracks and analyses WHO guidelines and strategies, including environmental toxins, heavy metals reduction and epidemic preparedness, with a focus on Ghana and Africa.*

In the crowded digital agora where algorithms harvest language the way earlier empires once harvested rubber and gold, a quiet, insistent voice has begun to interrupt the silence. Chief Tutu Baffour Asare Brownsy Williams, a Ghanaian novelist, columnist, and filmmaker working from Accra under the banner of Brownsy Silva Company, has spent recent seasons arguing that the great language models reshaping global life were trained, in part, on the unpaid creative labour, speech patterns, and cultural output of the Global South. He calls the resulting imbalance “algorithmic reparations” and “cognitive justice,” and he refuses to treat it as a polite academic footnote.

Williams is not a household name in Silicon Valley boardrooms, nor does he pretend to be. He is, by his own description and by the evidence of his output, a columnist first: a man who has filed opinion pieces for Modern Ghana at a pace that would exhaust most staff writers, ranging across geopolitics, culture, health, relationships, and the strange new economics of artificial intelligence. Beneath the journalism sits a larger architectural project. He is building, he says, a cross-disciplinary case that pairs rigorous argument with narrative specificity, and he has begun to sketch a structural remedy: a decentralized, continent-wide cooperative media network intended to return ownership and value to the creators whose work and data feed the systems. The proposal is deliberately ambitious. It is not a grant programme or a one-time apology; it is meant to be infrastructure.

Born and raised in Ghana, Williams came to storytelling early. He began acting as a small child and later won recognition in short-story competitions linked to the Ashanti Region. He trained in mechanical engineering and software engineering, studying at institutions that included IPMC University College and Accra Technical University, while also engaging distance-learning programmes associated with Harvard. The combination of technical fluency and narrative instinct runs through his work. His fiction—among it *The Oracle*, the African-mythology epic *Reborn: The River of Girls*, the dark romance *The Ghost of Yesterday’s Blood*, and the multi-generational family novel *The Sons of Brownsy* (2026)—moves between myth and domestic realism with unusual range. *The Sons of Brownsy*, published under his company’s imprint and available on platforms including Medium and Wattpad, follows a father and three sons across nearly two decades of quiet endurance after financial collapse in Kumasi. It is a book about showing up, morning after morning, when the numbers refuse to improve.

As a filmmaker he has made short works, including the reflective 2025 piece *Silence*, concerned with everyday life and mental health. Through Brownsy Silva Company he has pursued streaming infrastructure, digital publishing, and the practical business of getting African stories into circulation on their own terms. Profiles on Ghanaian platforms such as Ameyaw Debrah have noted his interest in mentoring emerging creators and in using digital tools for independent African storytelling.

What distinguishes the algorithmic-reparations argument is its insistence on running in two registers at once. Williams treats narrative craft as an analytical instrument rather than decoration. Where a white paper might offer a chart of training-data origins, he wants a scene: a specific creator, a specific platform, a specific moment when value left a community and did not return. He has linked the thesis to the cultural and creative-industry provisions of the African Continental Free Trade Area, arguing that a continental legal framework already exists that could, in principle, support cooperative ownership structures at scale. Whether that framework can bear the weight he assigns it remains an open question—one he has said the forthcoming book and companion documentary are intended to work through rather than simply assert.

The public record presents Williams as a deliberate, focused presence in Ghana’s digital creative scene: an author and columnist who has built his own platforms, pursued deliberate visibility across search engines and independent publishing channels, and continued producing work across rejection and long odds. Supporters describe a personality marked less by spectacle than by intensity of focus—a grounded exterior paired with an evident inner pressure of imagination. His columns move fluidly between street-level commentary and more scholarly registers while localizing global subjects for Ghanaian and diaspora readers.

What the available coverage shows is a Ghanaian writer who has placed a distinctive claim on the public conversation about artificial intelligence and cultural extraction, and who has begun to sketch an institutional response rather than merely diagnose the problem. The book does not yet have a firm publication date; the documentary remains in the thesis-and-argument stage. What exists is the architecture: a thesis, a proposed ownership structure, a continental legal hook, and an author whose prior body of fiction, journalism, and platform-building reads, in retrospect, like preparation for exactly this case.

Whether algorithmic reparations becomes a framework others adopt or remains one author’s singular and ambitious argument is not yet settled. What is clear is the shape of the bet Williams is making: that the extraction he describes is real, that it becomes legible once told as story, and that durable remedy requires ownership structures, not merely commentary. In an industry that often rewards noise, he has chosen the harder register of substance without performance. The ledger, he insists, remains open.Column: Tutu Baffour Brownsy Williams

Disclaimer: "The views expressed in this article are the author’s own and do not necessarily reflect ModernGhana official position. ModernGhana will not be responsible or liable for any inaccurate or incorrect statements in the contributions or columns here." Follow our WhatsApp channel for meaningful stories picked for your day.

Comments