For most of modern history, evidence had a physical bias. A document could be forged, yes. A witness could lie, of course. But a photo, a video, or a voice recording still carried a kind of intuitive authority. You could argue about context, but the underlying artifact usually felt real.
That assumption is now breaking faster than most institutions understand.
We are entering a world where visual and audio evidence is no longer self-authenticating. A CEO can be cloned. A politician can be staged. A market-moving announcement can be fabricated in a way that looks convincing for the exact ten minutes required to do damage. In cybersecurity, ten minutes is a lifetime. In financial markets, it is enough to move millions. In geopolitics, it is enough to trigger a response that cannot be taken back.
This is why I think the deep fake conversation is still too shallow. People frame it as a media problem, or a social platform problem, or a content moderation problem. It is bigger than that. It is becoming an infrastructure problem.
Because once “seeing is believing” dies, every digital system that depends on trust has to be redesigned.
We Are Not Losing Content. We Are Losing Default Trust.
The important shift is not that fake content exists. Fake content has always existed. Edited images, manipulated clips, staged narratives-these are old tactics with new tooling.
The real shift is cost. What used to require a studio, a budget, and real expertise can now be done by a laptop, a model API, and intent. The barrier to entry collapsed. And when the cost of fabrication collapses, the burden of verification explodes.
That changes the economics of every trust system online.
Until now, most platforms have operated on an implicit model: publish first, verify later, if ever. That was already fragile in the age of screenshots and virality. In the age of synthetic media, it becomes reckless. If anyone can generate a realistic executive briefing, earnings clip, support call, or incident video, then the old moderation model is permanently behind the attack curve.
Attackers do not need perfect deception. They need plausible deception delivered at the right moment, to the right audience, with just enough realism to trigger action.
That is a very low bar.
Cybersecurity Should Be Leading This Conversation
I find it striking that many of the strongest ideas in content authenticity are still discussed mostly by media, policy, and consumer tech circles. This is a security topic.
Why? Because synthetic content is becoming an attack vector.
A convincing fake voice note can trigger a fraudulent wire transfer. A fake board message can create internal panic. A fabricated incident screenshot can spread misinformation during an outage. A manipulated “executive directive” can bypass social friction inside an organization that already runs on speed and trust.
We spent years teaching companies not to click suspicious links. Now we need to teach them not to trust suspicious reality.
That is a much harder transition.
Traditional phishing exploited interface trust. Deep fakes exploit human trust more directly. They collapse the remaining gap between “this looks real” and “this must be real.” And that makes them especially dangerous in operational environments where people are trained to move quickly under pressure.
From a defensive perspective, this means authenticity can no longer live only at the content layer. It has to move into the control plane.
The Future Is Provenance, Not Detection Alone
Most public discussion still centers on detection: can we build models that identify AI-generated images, voices, or video? The answer is yes-for a while. But detection alone is a losing game.
We know this pattern from security. Pure signature-based defense works until adversaries adapt. Then you end up with a permanent cat-and-mouse cycle where defenders need high confidence and attackers need one bypass. Synthetic media detection will follow the same path.
That is why I am much more interested in provenance than in classification.
Instead of asking, “Does this look fake?” we should increasingly ask, “Can this artifact prove where it came from, when it was created, what modified it, and whether that chain is intact?”
This is a stronger question. It is also a more operational one.
Cryptographic signing, hardware-backed capture attestations, tamper-evident metadata, and verifiable publishing chains are not glamorous topics. But neither is TLS. Neither is DNSSEC. Neither are certificate revocation lists. Trust infrastructure usually looks boring right until the moment society realizes it is indispensable.
That is where content authenticity is heading.
Origin proof: Who created this artifact, using which identity?
Capture integrity: Was this recorded on a trusted device or generated synthetically?
Modification history: What edits happened after creation?
Distribution integrity: Has the artifact been altered between source and viewer?
Policy context: Should this artifact be treated as evidence, marketing, satire, internal communication, or untrusted user content?
That stack is much closer to infrastructure engineering than to social media moderation.
The Missing Layer: Operational Adoption
The frustrating part is that many of the building blocks already exist. Cryptographic signatures are not new. Device identity is not new. Secure enclaves are not new. Content credentials frameworks are emerging. Watermarking has improved. Standards work is happening.
And still, adoption lags.
Why? Because authenticity is a classic coordination problem. The value compounds only when the chain is end to end. If the camera signs but the editing tool strips metadata, trust breaks. If the publisher preserves provenance but the platform ignores it, trust breaks. If the platform displays a badge but the viewer does not understand it, trust breaks. If enterprises do not define policy for trusted versus untrusted media, trust breaks again.
In other words: the technology is not enough. We need operational discipline around it.
This is where I think the next wave of product design will matter. The winners will not be the companies with the cleverest detector demo. They will be the ones that make authenticity legible and actionable inside real workflows.
A newsroom needs provenance signals in editorial review. An enterprise needs provenance signals in executive communications. A security team needs provenance signals in incident response tooling. A bank needs provenance signals in approval flows. A messaging platform needs provenance signals at the point of consumption, not buried in a hidden menu.
The common pattern is simple: authenticity has to become usable.
Why This Becomes a Board-Level Issue Faster Than People Expect
There is also a leadership mistake here that many companies are making. They still treat synthetic media risk as a PR problem. It is not. It is a governance problem.
If your organization has no clear policy for how executives issue urgent requests, approve sensitive actions, or validate market-moving statements, then deep fakes do not need to fool the whole internet. They only need to fool one employee, one partner, one journalist, or one customer at the wrong moment.
That means boards and CEOs need to ask much sharper questions:
How do we authenticate executive communications in a crisis?
What channels are considered authoritative?
How do we handle voice-based approvals or urgent payment instructions?
Do our incident playbooks assume synthetic evidence may appear during an attack?
Can our employees distinguish verified communications from merely persuasive ones?
That is not paranoia. It is basic operational maturity for the next phase of the internet.
At Link11, we learned long ago that trust cannot be a soft concept when systems are under pressure. It has to be designed, encoded, rehearsed, and enforced. The same is now becoming true for content.
The Next Internet Will Need a Trust Envelope
I suspect we are moving toward a future where valuable digital artifacts carry something like a trust envelope: provenance, identity, integrity, and policy bundled together by default. Not because users will demand cryptography directly, but because every serious platform will eventually need a way to separate “asserted” reality from “verifiable” reality.
That distinction will matter everywhere:
In journalism, to establish source confidence.
In enterprise communication, to prevent social-engineering attacks.
In law, to assess evidentiary quality.
In finance, to reduce manipulation risk.
In consumer products, to preserve baseline trust in what people see and hear.
We should be honest: this will not restore the old world. The age of naive trust is over. People will become more skeptical, institutions more procedural, and high-value communication more tightly verified. That is not a cultural overreaction. It is an infrastructure adaptation.
And like every real infrastructure adaptation, the winners will be the ones who move before the failure becomes obvious to everyone else.
My Take
I do not think deep fakes will destroy truth. But I do think they will destroy convenience. The easy shortcut-the human instinct that a realistic artifact is probably genuine-is collapsing.
What replaces it will be slower, more cryptographic, and more operational.
That may sound cold. I think it is necessary.
Because in the next internet, trust will not come from what content looks like.
It will come from whether the system around that content can prove it deserves to be believed.
Follow the journey
Subscribe to Lynk for daily insights on AI strategy, cybersecurity, and building in the age of AI.
Subscribe →