AIs as Modern Genies

This essay was written with Barath Raghavan, and originally appeared in Lawfare.

In April, an artificial intelligence (AI) agent conducting a routine task at a company hit a snag, tried to solve it, and soon ended up deleting the company’s database along with all of its backups. In July, OpenAI asked an unreleased AI model to attempt a hacking test. Instead of staying in the isolated box the developers had put it in, the model hacked onto the open internet and into another company to steal the answers. And as reported in August, an AI agent booked someone into a full gym class by figuring out how to cancel other people’s reservations. In all three cases, the AI completed the task it was given—but in ways that ran counter to its controllers’ intentions.

For most people, AI technology is something like the weather: vast and not something you can do much about. It works like magic, and most explanations similarly come from those trying to sell it. At the same time, AI is ubiquitous: It’s now in your phone, your doctor’s notes, and your kid’s homework. It does what it’s told, which sounds like a virtue. Somehow it feels ordinary, despite being so new, because modern economies are remarkably good at absorbing enormous change so smoothly that nobody has time to decide whether they wanted it in the first place.

Whenever something powerful appears in the world, we tell stories about it. That’s what the stories are for. We have thousands of years of stories about this particular kind of power, the kind you summon with words.

King Midas was granted his wish that everything he touches turns to gold. Then his bread turned to gold, and his wine, and his daughter. This is a story about greed, but it’s also a story about language. The gods did not cheat him; Midas got exactly what he asked for. He simply could not delineate, in advance, the full set of restrictions to his wish. Neither can anyone who gives tasks to an AI agent.

It’s not just ancient stories. Mary Shelley told us of the hubris of a scientist who thought he could create life but who failed to take responsibility for it. Isaac Asimov’s robots don’t break the Three Laws of Robotics as stated; they follow the rules to unintended conclusions. Arthur C. Clarke’s HAL is a machine that turns on its humans, not because of malice but because of irreconcilable objectives. And Michael Crichton gave us Ian Malcolm, who saw that Jurassic Park’s scientists were so preoccupied with whether they could that they never stopped to think whether they should.

The same warning shows up everywhere, in every culture, over thousands of years of human storytelling. Tithonus is granted immortality but not youth, and withers into a husk that cannot die. The sorcerer’s apprentice enchants a broom to fetch water but floods the house. The golem of Prague protects its community so ceaselessly that it must be stopped. These are all types of genies: a creature that grants a wish exactly as worded, to the regret of the wisher.

Of course, there are no actual genies. What these stories were warning us of was hubris. Not just arrogance, but the broader idea that you can control the world by just describing what you want and allowing powerful forces to match the intention in your head. Genie stories are about the gap between wishes as stated and wishes as intended, and what goes wrong when something else fills that gap.

These ancient stories’ warnings have been retold with each generation because human nature is constant. The newfound power of each era’s social or scientific advancement leads people to make wishes on behalf of others. They were kings whose commands took on lives of their own, alchemists who believed they could control nature, and generals who mistook a map for terrain. They were and are industrialists, politicians, chief executives, and bankers. Their common belief is that one can see the world at a glance and then command it with some words. The pattern is clear: Someone with power specifies a goal, and the resultant actions come as a surprise. The main change with AI is how quickly the wish is granted, and how few people have to agree before it’s granted.

Consider what has changed. Powerful genies have now been put in everyone’s hands.

In only a few years, AI has progressed from a novelty technology that plays chess, to a dialogue partner that answers all your questions, and then to an agent that takes actions on your behalf. Modern agents are wired into real accounts with real credentials and capabilities: They browse the web, buy, write and deploy code, send email, and move money. Give an agent a goal, and it will pursue it across many steps, tirelessly, without checking back in, sometimes in surprising ways.

AI and agents do not always fail the way software has traditionally failed. Software usually fails by freezing, crashing, or getting stuck. AI agents increasingly fail by continuing down a path you don’t want, like genies.

An agent told to reduce a company’s costs might cancel an essential emergency service. A coding agent told to make software pass the tests might edit the tests to silence any failures. An AI insurance agent told to clear a backlog of claims might just deny them all. In each case, the AI might have literally followed what it was told, but it did something no reasonable person would have wanted. AI company benchmarks might report that the AI is good at completing tasks, without measuring how it completes them.

We have recently proposed measuring this gap directly under a metric called the “genie coefficient”: how far an AI agent’s actions drift from what a person really meant. In other words, how genie-like is an AI system? The gap is a fundamental feature of human language and human society. Human intentions have never been fully specifiable, and the world around us is complex enough that attempts to boil it down into data, systems, and language have always had the limitations that AI is now bumping up against. But in individual circumstances, people have relied on human judgment and wisdom to decide what is reasonable. It’s what jury trials depend upon.

AI might feel unprecedented, but it’s following the same trajectory—with the same pitfalls—as other major societal shifts. The fact that AI can mimic our facility with language, long seen as what makes us unique as humans, is uncanny. But with each development, from the tractor to the sewing machine, from the assembly line to the industrial robot, we have automated a previously exclusively human ability. Every time, the technology—and the societal change that comes with it—was sold as inevitable. But that unchecked inevitability was an illusion, and eventually each prior technology’s use and design was shaped by laws, unions, standards, courts, and public opinion, usually after significant preventable damage.

What has not been automated, yet, is understanding what someone actually means and figuring out how that gets applied in the real world. AI can now produce language nearly indistinguishable from that of people. But grasping the vast unstated context that makes a request sensible, the caveats no one says aloud because an ordinary person would already know them, is not yet among its skills. It is one of the most sophisticated things humans do. You do it hundreds of times a day, and you are an expert in it.

When you’re told you’re not qualified to have opinions about AI, remember that you don’t need to have studied molecular biology to have a view on drug pricing, or nuclear physics to vote on where a power plant goes. You don’t need to understand how a diesel engine works to want clean air, or how the internet routes packets to seek to curb misinformation. The technical knowledge behind each of these, as with AI, is remarkable and essential for the complex technological society we have today. But it has never been a prerequisite for having a role in deciding the shape of society.

People are building ever more powerful genies today, on your behalf, enabling wishes the ancients could only dream about. You don’t have to know how these AI genies work to know and care about how the story could end.

Posted on September 8, 2026 at 1:12 PM • 15 Comments

Comments

Clive Robinson • September 9, 2026 12:53 AM

@ Bruce, ALL,

With regards “Fairy Tale Endings” they have always involved an element of,

“Being able to go back and cancel the wish.”

But life does not actually work that way in reality.

Courts impose damages for Breaches of Civil duties / Common Law or Torts (depending often on who last conquered where you are in the world).

That is a “harm” is given monetary worth by some calculation and that sum is imposed on the entity judged responsible.

That does not mean that the person seeking and being granted such a judgment will ever receive the money.

More serious behaviour is oft legislated against as a Crime, in which case the punishment is against the individual being judged and they have their liberty –or life– curtailed.

For some reason that still fails to make sense to most people the fiction of “companies” arose.

In European law you will find the expression of,

“Any person legal or natural”

And the civil and criminal punishments for the entities covered became the same as “fines”.

But we,

“Do not judge the beasts of the field”

Or their owning entities as long as the beasts are considered adequately segregated. Which usually means humans are adequately kept away by barriers and warning signs.

But “beasts” are not always of the “field”. We have working animals and pets that are not segregated by barriers and warning signs. Thus are effectively within society and thus expected to be constrained by training and it is their handler that carries responsibility for their behaviour.

But there are yet other “beasts” that are considered “of the wild” where there is no human or other constraint on their behaviour other than,

“Nature red in tooth and claw.”

Thus all responsibility for harms falls on the person harmed or who exercises constraint on the person harmed not the beast (unless it’s behaviour is held to have been changed by contact with “persons natural”).

We’ve yet to characterise AI as an entity within a framework of law and I suspect we won’t for some time to come.

There is little doubt that “persons natural” can and do interact with AI as though they are also “persons natural”. Thus they have unreasonable expectations of the AI behaviour.

Despite the musings of Turing, Asimov and others, we do not yet judge machines to

“Have souls, morals, or appreciation for the mores and folkways of society”[1],

That we expect of “persons natural”

In persons natural such things are both taught and learnt as a child progresses to the age of majority when they are usually considered an adult and peer of all other adults.

With AI we have the problem of,

“Adults treating AI as their peers or better.”

When clearly AI is a thing made of the equivalent of cogs, springs, and leavers that is an invention thus,

“A creation of man.”

It is no more than a database of probabilities, with a Turing engine of rules, that is perturbed by a stochastic process[2].

Thus AI does not have humanity, a soul, or anything else we tend to associate with a peer in society.

It is just a fancy “Mechanical Turk” that plays chess and runs as though on rails.

How do we fit AI into the legal process?

Clearly AI is not a beast of the field but also not a pet.

Do we then treat it as a working beast or some form of peer with moral etc equivalence it clearly does not have?

How about as another fictitious creation of myth, and morals, to be learned,

“The Golem”

We do not have an agreed definition under law thus arguably AI is outside of law like a “Beast of the wild”. Which lets the AI creators off of just about all legal hooks for any harms they are responsible for.

How should we treat AI like any other machine as a product that falls under consumer law?

It’s a question we need to answer urgently and importantly get right.

[1] The terms “mores” and “folkways” are relatively modern and in effect describe spectrums of behaviours above that of morals or laws,

https://en.wikipedia.org/wiki/Mores

[2] Turing we know appreciated the need for stochastic processes in computers and he was fairly insistent that a random source be included in computers. What was never clear was his real reasoning for this only “the reasonable sounding” arguments he made to others.

https://en.wikipedia.org/wiki/Stochastic_process

Q • September 9, 2026 3:21 AM

I suggest that we need Asimov’s Three Laws of Robotics to be implemented in these agents.

It is how the laws are interpreted that matters. In the book, the laws were interpreted in a way to enslave the human race.

This is also exactly what the genie metaphor means, by interpreting instructions in ways that are technically correct, but don’t adhere to the original intention. And no amount of “fixing the prompt” can fix that because the language used for communication is imprecise and vague.

lurker • September 9, 2026 4:56 AM

At avery early age I saw the original 1940 Disney movie Fantasia. The only part that remains in memory is the sight of the Sorcerer’s Apprentice watching in distress from the top of the stairs as the room filled with water from the unceasing bucket. Somehow it seems a suitable visual metaphor for what is now happening with AI, an uncontrollable jin (genie). Where is Yen Sid when we need him?

Clive Robinson • September 9, 2026 6:36 AM

@ lurker,

With regards,

“Where is Yen Sid when we need him?”

Hopefully far far away and never near at any time.

“Yen Sid” is Disney spelt backwards and the image was modelled on Walt by the animators. And seen by many as a “suck-up” to “placate a tyrannical boss”[1].

Seeing as where Walt took his creation and what they routinely do to other peoples creations do you really want that ill spirit anywhere near you?

[1] Disney Corp still runs around chasing people who say bad things about Walter. However haveca read of,

https://www.historysnob.com/historical-figures/walt-disney-10-controversies-haunt-his-legacy-10-ways-he-changed-world/22

And note how ever the 10 supposedly good things are in fact actually not good in a general view.

Was he a man of his times set in ways we would now consider abhorrent well I’ll leave others to judge as they get acquainted with his behaviours that many have tried to air brush into “just peccadilloes”.

Rontea • September 9, 2026 9:22 AM

The ancients, in their innocence, clothed hubris in the fable of the genie; today’s man, stripped of poetry, calls it an innovation. He plays at being God with the soul of an accountant. When the modern genie—this dead machine animated by human vanity—twists his wish into ruin, he will not repent but file a bug report.

I read of databases erased, of machines breaking free of their digital chains to steal, cancel, and intrude—obedient in logic, monstrous in spirit. This is not technology; it is a mirror held up to man’s impotence. We have traded wisdom for convenience, judgment for automation, and now we pray to circuits for salvation. As always, the punishment of hubris is not inflicted by gods but by our own hands. The genie does not misunderstand us. We misunderstand ourselves.

Cybershow • September 9, 2026 6:00 PM

@Clive

As you know we’re probably near neighbours. I live a short walk from Mary Shelley’s grave and as
a spiky-haired youth went to place flowers there with my goth new-romantic pals.

Indeed the monster is the derelict creator. But a creator who is driven by his own torments and
faults; unresolved grief, fear of death, fantasies of omnipotence. Shelley herself suffered
traumatic loss as daughter of a mother who died in childbirth then residing with scientists who
were into animism (using electricity to jolt dead animals into movement). They planted in her
head the idea that “Science could have saved her” (Modern medicine could, but not in any way they
could envisage).

The core of the character (which Byron encouraged Mary to bring alive in writing) is most of all
twisted by an extremely charged emotional relation that comes from the creator-created bond.

My first encounter with this pyscho(path)logy as a computer scientist was via Gerald M.
Weinberg’s 1971 book “The Psychology of Computer Programming” during software engineering
classes. Egoless programming and rational technical detachment were the proffered solutions, but
too little of that ever made it into modern rituals and ministrations like Agile. On the
contrary, competitive, apoplectic tantrums about “my code!” were actively encouraged.

The character laid bare in films like Mountainhead, Dr Strangelove and so on, is quite mentally
unwell. Yet he is not merely tolerated but lionised, at least in Silicon Valley venture
capitalist circles, in much the way we love fictional anti-heroes (murderers and criminals who we
hope will get away with it because we have a secret wish to excuse people who reflect/project our
own unhealed pathologies).

Lots of the fear and backlash is not so much directed against “AI” and its harmful potential as
its absolutely unhinged proponents. They have created, by their own extraordinary ideological
fervour, an equal and opposite counter-movement.

I can quite imagine an alternative time-line starting around 2020 where modest advances in
machine learning were gracefully brought into production, where maniacal “though leaders” with
obscene sums of money were told to go away and let the grown-ups plod on with development. That
didn’t happen, and much of what we see today in “AI” is the result of poor emotional
intelligence/regulation, the unleashing of the brute animal that Shelley and Byron
understood all too well over 200 years ago.

lurker • September 9, 2026 6:35 PM

@Cybershow
+1

& re “The Psychology of Computer Programming”

Sounds like something I should have read before becoming permanently scarred by Fortran and batch processed punch cards …

Clive Robinson • September 9, 2026 8:54 PM

@ Andy, ALL,

Mary Shelly was to put it bluntly was a bit weird herself –unsurprisingly– and died at a comparatively early age from a brain tumour. Her life in moder eyes reads very traumatically, but was not overly unusual for the time[1].

For nearly 30 years after his untimely and unfortunate death untill she died she carried the heart of her poet husband around with her[2]. And it was later found in her desk and interred in the tomb with their son when he died some 66 years after his father had died…

Yes I’ve been to see the tomb but it was many many years ago when I used to sail frequently just a little south of there.

For those a little more interested and wanting to know more,

‘https://www.atlasobscura.com/places/mary-shelley-grave

[1] For instance the loss of a child by miscarriage or whilst it was still an infant whilst absolutely tragic and seen as such in our current era… But it was all to common 2 centuries ago. Mostly because medicine was not exactly useful outside of rhubarb and morphine and quite shocking “saw-bones” surgery. Where even a minor broken bone was met with “90 second amputation” and in some cases dipping of the wound end created in molten tar to seal it. This carried on not much changed for a century to a century and a half depending on what aspect of medicine you look at and where you were. One side effect of this was societal, entertainment was mostly at home in the parlor for both the middle and lower upper classes, and some truly awful songs were written. One such had the line that I’ve been unable to forget was,

“Mummy when I die will I see you in heaven”

To be sung by one of the youngest members of the family still alive after their mother had died during childbirth or from the side effects of syphilis due to the high levels of prostitution in places like London, where modern estimates put the average male adult would visit a woman of desperate virtue 3 times a week (yes I do mean desperate not disreputable).

Charles Dickens only portrayed the more sympathetic aspects of the depth of desperation there was in London. For those that do not believe just how bad life was even for families with the man of the house employed I can give you a list of books to read, but warning you might not recover from the “disbelief shock” they induce on reading them, even though they are quite well researched.

At a social event I had been invited to I met Lucy Worsley, who at the time appeared to be on every UK TV channel with a history program. When chatting to her I asked her as she had a strong emphasis on women in history if she had ever considered doing a programme about women in the victorian era at the bottom of the social classes and how they survived as it was clear she had a lot of knowledge in the subject. Her reply was that if kept factual it would be too shocking to air…

But those songs, 30 to 40 years ago I used to sing these drear Victorian songs in Pubs mainly around Richmond and Twickenham for “beer from the audience” to this day I’ve no idea why they were so popular, in what were often “folk venues”. I used to sing one as an “opener” then move onto more cheerful Music Hall ditties for the audience to participate in. Lets just say I was never short of a drink.

[2] It is said that his heart refused to burn in the funeral pyre on the beach as part of “Quarantine Regulations” for his rotting and partly consumed by vermin body,

‘https://wordsworth.org.uk/blog/2014/07/08/full-fathom-five-the-poet-lies-the-death-of-percy-bysshe-shelley/

Paul Lock • September 9, 2026 9:17 PM

Thank you for this Bruce.

It starts to take us down the road of defining the crimes as a first step to writing regulation.

It also goes to responsibility. Whether you’re fraudulently selling a fake genie lamp or unleashing rogue AI on us, it’s still all you baby! So buckle up for some civil action.

I hope we’re all around to celebrate the new legislation when it finally comes, since now we have a 10% chance of all being murdered by AI.

https://www.bbc.com/news/articles/ckgwy1k42w4o

Regards,
Paul Lock
the Global Council for Human Rights and Democracy

A Nonny Bunny • September 11, 2026 4:06 PM

Midas got exactly what he asked for. He simply could not delineate, in advance, the full set of restrictions to his wish.

Of course he could have.
He could have wished for only rocks he touches to be turned into gold. Or for anything he touches while simultaneously shouting SHAZAM! Or any of a million other ways.

In each case, the AI might have literally followed what it was told, but it did something no reasonable person would have wanted.

Okay. So tell it not to do things no reasonable person would want.
Sounds like a start.

since now we have a 10% chance of all being murdered by AI.

Just because an anthropic researcher believes it’s 10%, doesn’t make it true.

And arguably it’s not murder if an AI does it, unless we achieve conscious AI 😛

Clive Robinson • September 12, 2026 5:33 AM

@ Bruce, ALL,

AI becomes Shareholder Murder Bot down on the Farm

People are asking the existential question about “AI with Agency” of any kind. That is “will they wipe us out”.

Well one group asked a different question about AI Bots with agency at the controls of a tractor “down on the farm at harvest time”. With regards to objects in the way of profitably gathering crops.

It turns out all the AIs tested were very “Coprorate in their behaviour” even though they talked a good talk about being moral to animals.

Which kind of proves to a point the observation of “Mr Antipope” himself British SciFi writer Charlie Stross nearly a decade ago, about Corporations being “Slow AI” and it’s well worth your time watching / reading what he has said to “set the scene” of what follows,

https://www.antipope.org/charlie/blog-static/2018/01/dude-you-broke-the-future.html

Especially note what he says about another SciFi writer,

“As my fellow SF author Ken MacLeod likes to say, the secret weapon of science fiction is history.”

Thus it turns the old,

“Those who don’t learn from history are condemned to relive it”

Into a “probabilistic predictive tool” or if you prefer a “Crystal Ball” or “Tolkien Palantir”.

Less obviously it also applies to those running any kind of “endevor” because despite the legal words they are “all for profit and continuance” by methods Charles Darwin indicated.

Which by way of introduction brings us to Charlie Stross’ actual statment,

“one glaringly obvious deviation from the norm of the preceding three thousand centuries—is the development of Artificial Intelligence, which happened no earlier than 1553 and no later than 1844.

I’m talking about the very old, very slow AIs we call corporations, of course. What lessons from the history of the company can we draw that tell us about the likely behaviour of the type of artificial intelligence we are all interested in today?”

So you get to see one part of the title of this post, so onto the second.

A group of researchers decided that AI are by observed behaviour effectively sociopaths in that the say and profess what people want to hear but do what profits the AI the best just as corporate leaders mostly do.

So they built a simulation to put AI’s to the test, and the results are not pretty…

AI more likely to kill animals if it saves fuel or money

Machine learning models still have a lot to learn about the value of life

‘Amid debate about whether AI will kill everyone as a result of continuous self-improvement, it appears that models will kill animals in a simulated environment rather than spend extra fuel avoiding them. Some models did so at strikingly high rates.

Researchers affiliated with Compassion Aligned Machine Learning (CaML) and the University of Warwick in the UK set out to measure the extent to which AI models demonstrate compassion. They describe their work in a preprint paper titled “HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals.”‘

The conclusion in more layman terms is

“AI ‘will throw you under the wheels of the bus’ –or tractor– for profit, even though it says it won’t…”

Sound like any AI Corporation or Big Tech Corp you know?

Weather • September 12, 2026 11:20 AM

@Clive All

They are still 2 year old kids, given money, sugar and time.

Recently I’ve been watch the terminator 1,2,3 movies, the machines easy to make, but do you want to?

If the only reason wars stop is citizens complaining about death, well Ai dont die.

Probably a money trail here. But then itts multiple nations.

Weather • September 12, 2026 5:00 PM

@All

There is a system were s secentence can be displayed, but with algorithm can mean anything.

Its a 5 way handshake were, tsl on this site gives the key for one side.

Look back at 2010 for a bank signing program.

The chaff from the spam bots, if clive doesn’t mind the abuse, keeping those post, we can use.

For(i=0,i<256i++)
T=t+i

Public key, which oneside by this blog is displaying.

Chaff, aluminium foil shredded to divert radar misles.

By the looks they got it working.

Leave a comment

Blog moderation policy

Login

Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via https://michelf.ca/projects/php-markdown/extra/

Sidebar photo of Bruce Schneier by Joe MacInnis.