Why the "deepfake detection" strategy is not sustainable

AI-generated content accelerates faster than AI detection

In just a few years, deepfake technology has moved from "visibly fake" to producing videos nearly indistinguishable to the naked eye. Next-generation AI models can synchronize lip movements, simulate facial expressions, lighting, and voice with very high realism even at 4K quality. The telltale signs once used to identify deepfakes unusual eye movement, slight facial distortion, or lighting inconsistencies are gradually disappearing as AI models continuously improve.

The problem is that this is an asymmetric race. The deepfake creator only needs one convincing video to cause harm, while the detection side must accurately identify every piece of fake content within the enormous volume of data uploaded every day.

As AI-generated content continues to accelerate, a strategy that relies solely on "detecting fake videos" will become increasingly costly, slow, and incapable of achieving perfect accuracy. This is also why many countries are beginning to shift toward a different approach: verifying the origin of authentic content at the point of publication.

Ordinary users have almost no verification tools

Imagine someone receiving a video call from a "family member" in a panic over an accident, urgently requesting a money transfer. In the few short minutes available to make a decision, most users have no way to verify whether that video is real or a deepfake.

Current deepfake detection tools have yet to achieve perfect accuracy, typically require a degree of technical skill, and are almost entirely unsuited to real-time situations. For ordinary users, "distinguishing real from fake by instinct" becomes the last line of defense.

That is also the most dangerous point: when digital trust depends on personal intuition rather than a standardized verification mechanism. In an environment where AI can fake faces, voices, and expressions with growing realism, people will find it increasingly difficult to protect themselves without a trusted content provenance verification infrastructure.

Shifting the focus: from "detecting fakes" to "verifying authenticity"

An approach being championed by many international organizations is to stop trying to detect every piece of fake content, and instead build a mechanism to verify the origin of authentic content at the point of publication.

In this model, each video or image can be accompanied by the creator's digital signature, information about the recording device, the publication timestamp, and the edit history. Viewers no longer need to guess whether content "seems real or fake”, they can directly check whether the content has a verifiable origin.

This is also the principle behind initiatives such as C2PA (Coalition for Content Provenance and Authenticity), co-founded by Adobe, Microsoft, BBC, Sony, Intel, and many other major technology organizations to establish a standard for verifying the origin of digital content.

In this context, national blockchain can serve as a trusted infrastructure layer, where digital identities, digital signatures, and content authentication proofs are stored and independently verified. Rather than chasing every new deepfake that appears, the system shifts to verifying which content is authentic, who published it, and whether it has been altered without authorization.

🔑 Read more: Decentralized identity as a global trend in the digital era

The role of national blockchain in content provenance verification

Every content creator - a verifiable digital identity

In a national blockchain model based on DID, every individual or organization publishing digital content can hold an identity in the form of did:nda:.... This could be a journalist, photographer, KOL, enterprise, government agency, or any content creator who needs to verify their digital identity.

Each article, image, or video published can be digitally signed using the private key linked to that DID. This allows viewers to immediately check three important factors:

• Who published the content

• Whether that identity has been verified

• Whether the content has been altered after signing

The entire verification process takes place within seconds through the blockchain infrastructure, without relying on subjective trust or a single intermediary platform.

For example, if a video surfaces claiming to show a business leader making statements about strategy changes or share prices, viewers can check whether the video carries a digital signature linked to the spokesperson's official registered DID. If no valid authentication is present, the system can immediately flag the content as having an unverified origin.

In the deepfake era, value no longer lies in whether something "looks real", it lies in the ability to prove that the content genuinely came from a specific source.

Government agencies, media outlets, enterprises - issuers of digital trust

In the national blockchain model, government agencies, media organizations, and major enterprises can serve as Issuers — entities that publish digitally verifiable credentials that can be publicly checked.

For example:

• The Ministry of Information and Communications confirms the official DID of a media outlet

• A newsroom issues a VC verifying that a journalist is a staff member or authorized contributor

• An enterprise confirms the DID of its official spokesperson

• A government agency digitally signs its public announcements and statements

As a result, when a video or article spreads with claims like "this is a statement from Minister X" or "an official announcement from Company Y," viewers can immediately check whether the content was signed by the officially registered DID.

This mechanism not only helps reduce the risk of deepfakes, but also addresses a broader challenge: verifying the origin of information in a digital environment. In an era of mass AI-generated content, the ability to know "who actually published this" will become a trust layer as important as the content itself.

Verification at the point of capture

An important development being tested by Sony, Canon, and Nikon is integrating a data verification mechanism directly into the recording device. Each image or video is digitally signed at the moment of capture, along with metadata such as timestamp, location, and device information.

This authentication record can be cross-checked through blockchain infrastructure to prove that the content is the original and has not been altered after publication. If a video is edited, effects are added, or AI intervention occurs, the original digital signature will no longer be valid.

In sensitive fields such as investigative journalism, political media, security surveillance, or legal evidence, this mechanism creates a "digital original" that can be independently verified and carries significantly greater legal value than an ordinary digital file.

For this model to operate at scale, coordination is needed between device manufacturers and national verification infrastructure, which receives, traces, and verifies the digital signatures of content from the moment it is created.

🔑 Read more: Vietnam's National Blockchain and the Foundation of Trust Infrastructure in the Data Economy Era

4 specific deployment scenarios in Vietnam

1. Digital spokesperson portal for public agencies

One of the biggest weaknesses in today's digital media environment is that misinformation typically spreads faster than verified information. Within minutes of an event, social media can be flooded with alleged "leadership statements," "urgent announcements," or videos purportedly from government agencies — while official bodies need additional time to verify and respond.

That gap is the ideal environment for deepfakes and misinformation to thrive.

With national blockchain infrastructure, each agency can operate an official digital spokesperson portal based on DID. Every announcement, video, or public statement is digitally signed using the agency's identity and published as a Verifiable Credential or content with verified provenance.

Citizens, media, and digital platforms can immediately check:

• Was this content actually published by that agency?

• Does the DID that signed the content belong to a registered agency?

• Has the content been altered since it was published?

In this model, those spreading misinformation cannot fake the publishing source without possessing the agency's valid signing key. This helps close the trust gap between "content that spreads" and "verified information" in the digital environment.

🔑 Read more: Why is Vietnam building a national blockchain platform?

2. Digital journalism and verifiable content

In an information environment where content is constantly clipped, recontextualized, and shared at speed, media outlets not only need to publish quickly — they also need the ability to prove the origin and integrity of their content.

With national blockchain infrastructure, each article, image, or video can be linked to a cryptographic proof at the time of publication: which outlet published it, when, what the original version was, and whether the content was subsequently altered.

When an article is reshared across multiple platforms, readers can still verify its original source through the newsroom's DID and digital signature. If a correction or update is made, the new version is also digitally signed and linked to the previous version, making the full edit history transparent and traceable.

For investigative journalism in particular, where content may face the risk of deletion, alteration, or forgery, blockchain creates an immutable tracing layer that protects the integrity of the work and the evidentiary value of the reporting.

3. KOLs, influencers, and advertising with verified provenance

Deepfakes are fueling a surge in fake celebrity endorsements on social media: from videos of "singers promoting medicine" and "actors investing in financial products" to cloned voices used for sales scams and fraud.

In the DID model, each KOL or public figure can hold a verified digital identity. Official advertising content would be digitally signed with that DID before publication.

Users and platforms can immediately check:

• Was this video actually published by that public figure?

• Has the content been altered after signing?

• Does the advertisement come from a verified official account?

If no valid signature from a verified DID is present, platforms can flag or refuse to distribute the content.

This mechanism not only protects consumers from fake advertisements, but also safeguards the reputation and digital brand assets of influencers themselves in an environment where AI-generated content is becoming increasingly widespread.

🔑 Read more: Vietnam's data economy 2026: The strategic role of national blockchain

4. Legal evidence and surveillance video

Video from security cameras, dashcams, and mobile phones is becoming an important source of evidence in many civil and criminal cases. However, in an era of AI video editing, the question "has this video been tampered with?" is becoming increasingly difficult to answer.

A new approach involves digitally signing content at the recording device and anchoring the authentication proof on the national blockchain. Each video created would come with a timestamp, device information, and an immutable digital signature.

If the video is subsequently altered, the signature becomes invalid. This creates a "digital original" that can be independently verified and carries significantly greater legal value than an ordinary video file.

Estonia, the UK, and South Korea have begun piloting similar models in the justice and security sectors. With national blockchain infrastructure and a legal framework covering personal data and electronic transactions, Vietnam is in a position to progressively deploy digital evidence verification use cases along these lines.

🔑 Read more: Competitive Advantage in the Data Economy Era

Digital trust is a public asset

Fighting deepfakes is not a problem that any individual or platform can solve alone. A social media platform can moderate content within its own ecosystem, but cannot control the entire internet. A user may hold a DID and digital signature, but if the surrounding platforms and organizations do not support verification, the value of that identity is limited.

This points to an important reality: digital trust is a form of public asset. It only delivers value when many parties participate in the same verification infrastructure and adhere to the same common standards.

This is also the role of the national blockchain: to create a shared trust layer so that government agencies, enterprises, media outlets, digital platforms, and citizens can verify the origin of content through the same mechanism, rather than each party building its own isolated "verification island."

With the did:nda method registered in the W3C DID Method Registry and verifiable through the DID Universal Resolver, NDAChain is designed so that digital identities and content published from Vietnam can still interoperate with W3C-compliant international ecosystems.

🔑 Read more: NDAChain positions Vietnam among Asia's pioneering nations in decentralized identity under the W3C DID standard

In a world where digital content crosses borders in seconds, digital trust also needs the same capacity for global interoperability.

Challenges to address

Verifying content provenance through national blockchain and digital identity is a promising direction, but deploying it at social scale requires simultaneously addressing technical, legal, and ecosystem challenges.

Balancing verification with the right to anonymity

Not every content creator can publicly disclose their real identity. Investigative journalists, whistleblowers, and civil society groups sometimes need to have their personal identity protected.

The system therefore cannot be designed on the principle that "every DID must reveal a real name." A more appropriate approach is to allow DIDs to be verified through a sponsoring organization or trusted intermediary platform, so that content still has verifiable provenance without requiring the public disclosure of personal identity.

Impact only comes when the ecosystem is large enough

Content verification cannot be effective if only a handful of agencies or individuals participate. The system simultaneously needs:

• Enough official content publishers signing their content digitally

• Enough platforms supporting verification

• Enough users equipped with verification tools

This is a two-sided network problem: users will only verify when digitally signed content is widespread, while organizations will only sign content when a sufficiently large verification ecosystem exists.

Integration with international platforms is essential

Most digital content in Vietnam is currently consumed through global platforms such as Facebook, YouTube, TikTok, and X. A national content provenance verification system will therefore struggle to be effective without the ability to integrate and interoperate with these platforms.

This requires Vietnam to both develop its own infrastructure and participate in international standards and initiatives such as C2PA, W3C DID, and open provenance protocols, to ensure global compatibility.

The legal framework needs to lead

Many new use cases, such as digitally signed official statements, legally evidential video, and advertising with verified provenance still require specific legal guidance.

Regulations covering electronic identity, the legal validity of digitally authenticated content, and the responsibilities of distribution platforms will be the foundation that allows the technology to operate stably at social scale.

The deployment roadmap must be phased

Rather than rolling out broadly from the start, the appropriate approach is to begin with high-value use cases where deployment is clearly feasible, such as:

• Digital spokesperson portals for public agencies

• Digital journalism with verified provenance signing

• KOL and advertising content with verified DIDs

Once a verification habit and supporting ecosystem have been established, the system can expand to more complex challenges such as legal evidence, real-time video authentication, and large-scale AI-generated content.

Blockchain and the reshaping of digital trust

Deepfakes are not merely a technology problem, they are a direct challenge to social trust in the AI era. When images, voices, and videos can all be fabricated at near-zero cost, the greatest risk lies not just in individual fraud cases, but in people gradually losing the ability to trust what they see.

🔑 Read more: Building digital trust: The foundation for a prosperous Vietnam

In this context, national blockchain is not the only solution, but it is an important infrastructure layer for verifying content provenance and the digital identity of publishers. Rather than simply chasing fake content, the new approach focuses on the ability to prove which content is authentic, who published it, and whether it has been altered.

NDAChain is designed as a trust infrastructure where every piece of trustworthy content can have its origin verified, every official statement carries an authenticated signature, and every citizen has an additional tool to protect themselves in an increasingly complex digital information environment. In the battle between AI-generated content and social trust, the ability to verify will become an infrastructure as important as the content itself.

👉 Learn more about NDAChain - Vietnam's National Blockchain Infrastructure at https://ndachain.vn/en