Schneier on Security

Clive Robinson • September 4, 2026 9:07 AM

@ Bruce, ALL,

Maybe it’s because I’ve been in effect away for a couple of months due to ill health and more recently for a couple of weeks due to being in hospital in isolation without communication that I find myself coming back to quite a change with regards AI even here.

I’ve always sounded a cautionary note with “Current AI LLM and ML Systems” and have not been down the all to obvious Microsoft and Co “Be Business Plan” to,

“Surveillance on individuals via built in AI.”

The first step of the “Be Business Plan” being “Bedazzle,” and the last “Betrayal”.

The Microsoft aim being clearly to copy via the Internet to their “cloud” everything a user does right down to spoken words and typing cadence.

Certainly enough to,

“Impersonate individuals beyond most humans ability to detect the impersonation.”

Which has significant identity and evidentiary concerns.

I’ve seen this as a major threat vector for sometime and have indicated as much. Also pointing out the “directing mind” and “arms length deniability” issues AI systems will give authoritarians and political actors.

Whilst the general trend was to talk up the “joys of AI” and it’s “possible mankind benefits”. It’s nolonger true.

Now I am sort of back I find that it appears,

“A switch has been flipped”

That all to suddenly my “caution” and that of the likes of Gary Markus have gained,

“Public Normalisation and Acceptance.”

It’s almost like I’ve become like

“Alice through the looking glass.”

Where the world has reversed it’s self.

Even the “Governor of the Bank of England” is “doom swiping AI” due to the risk on not just the UK but Global economy.

It’s as though the “spells cast” in the “Bewitch” phase of the MS AI BE Business plan have all failed.

With talk of,

“Genies escaping bottles”

And mentioning of the,

“Third wish that undoes the first two”

Being increasingly demonstrated.

It’s as though the world outside of ICT and Venture Capitalists gulling investor stupidity has broken through and brought reality with it.

Has

“The AI Hype Bubble Burst?”

I’d argue “not yet” but all of a sudden there is the chill of millions of “cold showers hitting” and people,

“Waking from enchanted sleep”

Thus a form of rationality rising to the fore.

So I hope for a couple of things,

1, The acceptance that AI has significant risk increases thus caution applies.
2, That people continue to view AI in a more realistic way as just a tool in the box.
3, That the AI hype bubble deflates slowly rather than suddenly implodes.

Much of this is happening quietly it appears as a response to China’s public facing behaviours with AI.

They have in effect “drained the moat”, and “demolished the runway” by “Robbing the wildfire burn-rate of oxygen”. Thus turning hype into ashes of the US AI companies AGI and similar claims.

So making “Narrow AI” increasingly the way to go. And leaving the US AI entities increasingly vulnerable.

I personally think even the “Current AI LLM and ML Systems” of the US Entities do have a limited future, not just for niche activities that AlphaFold etc have demonstrated but in slightly wider activities.

But the real future of LLM AI is in more general ways at a much more personal level using,

“Small models as specific tools”.

That is what started as,

“Highly expensive vibe coding and uncontrolled agentic acts, on systems so large they failed.”

Is reigned in and run privately on commodity hardware for the likes of STEM related activities.

Which kind of implies that the old

“All trade roads lead East”

Has come about rapidly with the way China is presenting their public facing AI development.

Remember the old saws about,

“Jack of all trades master of none.”

And,

“Bigger is not better.”

Can now be seen to be true of,

“Current AI LLM and ML Systems done the US capitalist Way”

“Of stack it high, push it hard, untill something breaks.”

And will become increasingly clear is mostly the wrong way to go from now onwards.