When Seeing Is No Longer Believing: Elections in the Age of AI

For most of the internet's history, seeing was believing.

A video of a politician speaking was evidence they said something. A recording captured someone's voice. A photograph documented an event.

None of those things were ever perfect proof, but creating a convincing fake required time, skill and resources.

Artificial intelligence is changing that.

Today, realistic images, audio and video can be generated in minutes. As the technology improves, the difference between authentic and synthetic content will become increasingly difficult to recognize with our eyes alone.

For elections, that's becoming a serious problem.

Canada has already responded.

In June 2026, the Strong and Free Elections Act became law, introducing new protections against realistic deepfakes of electoral actors intended to mislead voters, among several other changes designed to address emerging threats to federal elections.

Legislation can punish the misuse of this technology.

But it raises a bigger question:

What happens when we can no longer rely on what we see as proof of what is real?

The Cost of Creating Something Fake Is Collapsing

Election misinformation isn't new.

Misleading advertisements, false statements, impersonation and propaganda existed long before artificial intelligence.

What's changing is the scale.

A convincing fake once required editing skills, equipment and significant effort.

Now, someone can potentially generate a realistic video of a candidate, imitate a voice, produce thousands of pieces of written content or create convincing images with tools available to almost anyone.

And the technology is only getting better. That changes the problem. The question is no longer simply whether fake content exists.

It's whether the average person will reliably be able to tell the difference.

Detection Won't Be Enough

The obvious response is to build better AI detection.

If AI creates something, another system should be able to detect it.

But that creates an ongoing race.

Generation improves.

Detection improves.

Generation adapts.

Detection adapts again.

There will absolutely be a role for detection technology, but there's another approach worth considering.

Instead of only asking:

"Can we determine whether this is fake?"

We can also ask:

"Can we prove where this came from?"

Those are very different questions.

From Detection to Verification

Imagine watching an important announcement from a candidate.

Instead of relying entirely on whether the video looks authentic, there could be cryptographic evidence attached to the content establishing where it originated and whether it had been altered.

That concept is already being developed.

Emerging standards for digital content provenance can attach cryptographically secured information to images, videos, audio and documents, creating a record of how content was created and modified.

It doesn't magically tell us whether something is true.

That's an important distinction.

But it can help answer questions like:

Where did this come from?

Has it been changed?

What happened to it before it reached me?

In a world filled with synthetic media, those answers become increasingly valuable.

Elections Have the Same Problem

This challenge extends beyond campaign videos.

At its core, an election is an information system.

Someone claims they're eligible to vote.

A voter submits a ballot.

A system records it.

Votes are counted.

Results are published.

Administrators certify the outcome.

At every stage, we're being asked to trust that information is authentic.

That's why the rise of AI makes verification much bigger than detecting deepfakes.

If creating convincing digital information becomes almost effortless, proving authentic digital information becomes increasingly important.

Verification Doesn't Mean Exposing Everything

There's an important balance here.

A more verifiable internet shouldn't automatically mean a less private one.

We don't necessarily need to know everything about the person behind a piece of information.

Sometimes we only need proof of one fact.

This person is an eligible voter.

This announcement came from the official campaign.

This ballot was successfully submitted.

This election record hasn't been changed.

This result came from the authorized system.

The challenge is building systems capable of proving those things without unnecessarily exposing the people behind them.

That's especially important in elections, where identity must sometimes be verified while ballot choices must remain private.

The Internet Was Built for Sharing. Now It Needs Proof.

For decades, we've focused on making information easier to create and distribute.

We succeeded.

Almost anyone can now publish something and reach thousands—or millions—of people instantly.

Artificial intelligence is accelerating that ability again.

But our systems for proving authenticity haven't evolved at the same speed.

That gap is going to matter.

Not just for elections.

For journalism.

For identity.

For financial transactions.

For education.

For communication.

For almost everything we do online.

The next major challenge of the internet may not be giving people access to more information.

It may be giving them a way to determine what deserves to be trusted.

Elections Are an Early Warning

Canada's decision to specifically address election deepfakes is one example of governments beginning to respond to this new reality.

There will be more.

New laws.

New detection systems.

New standards.

New verification technologies.

None of them will solve the problem alone.

But together, they point toward an important shift.

For a long time, digital trust was based largely on what we could see, who sent something and which platform it appeared on.

That may no longer be enough.

The internet is entering an era where seeing isn't necessarily believing.

Proof will have to become part of the experience.