AI-enabled glasses. The Laws Haven't Changed. The Assumptions Have

Like a lot of people, I've been watching the recent discussions around AI-enabled  glasses, wearable assistants, and increasingly capable cameras. 

One thing I've gradually come to appreciate is that laws don't just regulate behaviour. They quietly assume how the world works. 

Most of the time we never notice those assumptions because they're true. Recording someone requires a camera. 

Keeping information requires storing it. 

Recognising someone's face requires remembering what they look like. A photograph generally strengthens confidence that something happened. 

Those assumptions have sat quietly underneath privacy law, evidence law, and social  expectations for decades. 

Nobody talks much about them because nobody needs to. 

Or at least they didn't. 

Now we arrive at AI-enabled glasses. 

Curiously, the technology itself isn't what caught my attention. 

Nor is the politics. 

People far more qualified than me will continue debating whether these devices should  exist, whether they're socially acceptable, and whether regulators should step in. 

What interested me was something much quieter. 

For most of my life, when I saw someone wearing glasses, I made an unconscious  assumption. 

They were helping the wearer see. 

That assumption didn't need legislation. It most certainly didn't need guidance. It simply reflected what glasses had always been for.

AI glasses quietly change that. 

Now I don't necessarily know whether the person wearing them is simply looking at me... ...or asking an AI who I am. 

That's a very different world. 

For most of my career, (and life for that matter) recording another person was an obvious  act. 

You pointed a camera. You lifted a phone. You deliberately pressed a button. Even hidden cameras were still cameras. 

The law had something concrete to examine. 

• Intent. 

• Consent. 

• Location. 

• Reasonable expectations of privacy. 

Those ideas still matter. 

What has changed is that technology is beginning to separate concepts that used to arrive  together. 

An AI-enabled wearable doesn't necessarily "record" in the way we traditionally imagined. It might identify someone without saving a photograph. 

It might remember a conversation without storing the audio. 

It might recognise a face, infer an emotion, answer a question about the interaction, and  retain none of the original sensory data. 

That feels like a small distinction and I'm not convinced it is small. 

For years we quietly assumed observation and memory belonged together. If somebody remembered your face, they must have seen it. 

If they wanted to remember it later, they probably needed a photograph. If they wanted to search thousands of faces, they needed a database.

Those were separate activities. 

AI has started collapsing them together. 

Observation no longer necessarily leaves behind the evidence we expect. Memory no longer necessarily requires storage in a form humans recognise. Recognition no longer requires another human doing the recognising. 

Those aren't just technical changes, and the crux of my point is that they're changes to  assumptions that have quietly supported legislation for years. 

The same thing is happening elsewhere. 

Large language models separate authorship from writing. 

Voice cloning separates identity from speech. 

Deepfakes separate evidence from photography, or worse video. 

AI agents separate action from immediate human intent. 

Wearable AI separates observation from recording. 

Different technologies but the same pattern. 

The assumptions underneath yesterday's rules no longer fit today's behaviour quite as  neatly. 

"It's probably AI." 

It also made me think about evidence. 

For decades we've generally treated photographs and video as increasing confidence. Not perfectly because forgery has existed for centuries. 

Photographs have been manipulated for almost as long as photographs have existed. 

But when somebody produced a photograph, the conversation usually moved towards  believing it.

Now we seem to be entering a rather strange period where authentic evidence sometimes  has to defend itself against something quite brand new.  

"It's probably AI." 

I've heard variations of that sentence more often over the past year than I expected. That's an extraordinary shift. 

Technology isn't simply making fabricated evidence easier to create. 

It's changing how willing we are to trust genuine evidence. 

I suspect that psychological change may end up being more significant than the  technology that caused it. 

A different problem. 

People often describe AI as creating entirely new legal problems. 

I'm beginning to wonder whether that's the right way to think about it. 

Increasingly, I suspect AI is exposing assumptions that have been sitting quietly  underneath existing laws for decades. 

The laws themselves may not be broken. 

The assumptions they relied upon might be. 

That's a different problem. 

And, to me at least, a much more interesting one. 

One thing I've gradually learned about governance is that it usually begins before legislation  changes. 

Governance begins when somebody notices that reality has drifted away from the  assumptions written into yesterday's policy. 

Long before regulators publish guidance. 

Long before standards committees meet. 

Long before lawyers decide what the new boundaries should be.

Someone simply notices that the world has become stranger than the policy expected it to  be. 

That observation is often where governance starts. 

Moving forward.  

I’ll get to the moving backwards part in a moment. Believe me. 

I don't think AI is forcing us to invent entirely new principles. I do however think it's forcing  us to rediscover principles we stopped noticing because they had quietly become  assumptions. 

Assumptions about privacy. About evidence. Identity. Memory. Observation. "Memories fade. They were built that way for a reason." 

Years ago, I watched the movie Strange Days. There's a line that stayed with me long after  I'd forgotten most of the plot. 

"Memories fade. They were built that way for a reason." 

At the time I thought it was simply a memorable piece of dialogue. 

Looking back, I think it was describing something much deeper. Human memory isn't  perfect and it was never supposed to be. 

We forget names. We misremember conversations. We gradually lose the sharp edges of  unpleasant experiences. We often think of those as bugs, increasingly, I'm wondering  whether they're features. 

Our society evolved assuming human memory had limits. 

AI doesn't forget in quite the same way because it can observe continuously. It can retrieve  instantly. It can correlate experiences that a human being would never naturally connect. 

If enough of those systems begin surrounding us, then "forgetting" slowly changes from a  property of human memory into a design decision made by engineers. 

That's the part of Strange Days that I understand differently now. 

“They were built that way for a reason."

I don't know whether memory was built that way for a reason. I most definably do know that  our societies were. 

Our friendships, our relationships, our legal systems and our expectations of privacy all  emerged in a world where people forgot things. Not everything. Not immediately. But  gradually, imperfectly, and naturally,….. and so that our brains don’t shake themselves  apart trying.  

I suspect many people who have lived through the worst days of their lives are quietly  grateful that memory changes with time, even if it never disappears completely. 

AI isn't changing human memory. It's changing the environment that human memory  evolved within. For the first time, remembering everything is becoming technically possible. 

Which means forgetting is no longer simply something that happens. 

It's becoming something that has to be allowed. 

Wrapping this up.  

The difficult questions over the next decade may not be about what AI is allowed to do. They may be about which of yesterday's assumptions still deserve our confidence. 

Because sometimes the law hasn't changed at all. 

The world underneath it has. 

Perhaps the next generation of AI governance won't be judged by what our systems  remember, but by whether we still know when they should forget.