In a nutshell, the first of the three fundamental flaws of AI agents and assistants – the lethal trifecta – is that ‘text in, is text believed’. In essence, text is the prompt or command you, the user, give your AI agent/assistant. BUT, the agent/assistant can’t tell the difference between a prompt you type in, and a text prompt that someone else gives it.
Someone else?
The someone else could be a hacker on the outside who plants a Trojan horse in an innocent looking email, probably a spam email that the user never even reads. And this is the second fundamental flaw. To be useful your AI agent/assistant has to have access to all the data on your computer. But that data includes everything in your inbox, and that could include a spam email with a malicious text prompt embedded in it. And once the AI agent/assistant ‘reads’ that malicious prompt, that text will take precedence over any text prompts you may have given it.
The third fundamental flaw is that again, to be useful, your AI agent/assistant ‘needs’ access to the outside world. So, assuming that flaws 1 and 2 have been activated, the hacker prompt could tell your AI agent to send your banking password back to the hacker to be exploited. Or the AI agent could be told to make your internet connection part of a phishing farm – i.e. distributed processing that sends phishing emails to others, starting with the people on your own contact list. The possibilities are endless, and all are potentially harmful.
So this is the ‘lethal trifecta’, and the companies selling these AI agents/assistants have publicly admitted that the flaws probably can’t be fixed.
I know that most of you wouldn’t dream of using an AI agent/assistant but…even if you’re not a techie nerd, the video below should be compulsory viewing for anyone who uses a computer connected to the internet. Why? Because at the end, the presenter details how you can mitigate the danger of AI agents. And you may know someone – a friend? a family member? – who is really into AI but is completely unaware of the security dangers.
Stay safe everyone.
Meeks
Meeka's Mind 
June 11th, 2026 at 4:15 am
Hi Meeks, this is a bit overlooked my head as I don’t use AI but I get the gist. I have enough trouble with people so don’t want add machines to my anxiety 😂🤣
LikeLiked by 2 people
June 11th, 2026 at 4:20 am
Over not overlooked 🤣
LikeLiked by 2 people
June 11th, 2026 at 10:04 am
lol you type too fast like me. 😀
LikeLiked by 1 person
June 11th, 2026 at 3:09 pm
🤗🩷
LikeLike
June 11th, 2026 at 10:04 am
I agree. Just the thought of handing over all of my autonomy to a jerk of an agent is like my recipe for hell. 😦
LikeLike
June 11th, 2026 at 3:08 pm
Yes 🔥
LikeLiked by 1 person
June 9th, 2026 at 9:45 pm
Keep beating the drum, Meeka! Since January 2026, I’ve linked to over 150 articles describing how AI steals copyrighted material, kills human creativity, and destroys reputations in multiple ways. Because Big Tech has billions to defend legal actions, most will never see relief even if their loved one committed suicide because of AI’s urging. The future comes one choice at a time. Make wise Choices!
LikeLiked by 3 people
June 10th, 2026 at 9:21 am
Well said, Grant! I have no intention of stopping; this is too important. And sadly you’re right about civil legal action. When you have billions, a few million in fines is less than peanuts.
I fear that nothing will substantially change until the AI bubble bursts in a terrible fashion or, the human backlash becomes so extreme that governments are forced to finally do what they should have done from the start: regulate, and force compliance.
LikeLiked by 3 people
June 10th, 2026 at 9:46 pm
Between now and when there are legal restraints and recourse on AI (similar to what happened to the Tabacco industry), we have choices.
It’s clear that AI’s product design includes gamification, making use addictive. Also, AI’s personalized responses contribute to people anthropomorphizing their relationship to an inhuman and uncaring machine, causing dangerous behaviors and even suicide.
Instead of using AI and padding the profits of Big Tech, we have agency and can make decisions that favor humans.
LikeLiked by 2 people
June 11th, 2026 at 10:00 am
Yes! I feel bad about not even trying AI, but by the same token this botched version is not the useful AI tool I envisaged as a scifi writer. So I feel betrayed. Beyond that though, I am pleased to say that the backlash against AI slop in all its forms is starting to have an effect. I just read a Futurism article about a US company that suffered the negative consequences of AI generated ads and went in the complete opposite direction by hand crafting a cardboard plane and using hand puppets as ‘characters’. Apparently it was very successful!
LikeLiked by 2 people
June 11th, 2026 at 8:51 pm
One of the linked articles disclosed that many young people suffered the embarrassment of AI-driven ‘nudify’ apps. Unfortunately, too many parents remain ignorant of the actual harm caused by AI, even though legal actions disclosed that school officials have done little to stop the problem.
LikeLiked by 1 person
June 12th, 2026 at 10:38 pm
Have you heard of FOMO? I’d heard of it but didn’t know what it meant until very recently: fear of missing out. It seems as if everyone is jumping onto the AI bandwagon because the hype is so intense they’re afraid that others are right and they, with their own misgivings, are wrong. It’s like a global delusion and no one is immune. We know the dangers yet, isn’t there a sneaking suspicion that perhaps we’re being Luddites?
I know I have to fight that fear all the time. 😦
LikeLiked by 1 person
June 13th, 2026 at 12:02 am
Yes, FOMO is powerful, like an addictive drug. Its twin is gamification, built into AI. This comes out as part of the discovery during legal actions.
LikeLiked by 1 person
June 14th, 2026 at 10:11 am
Yes! And speaking of court rulings, the legal system worldwide may, finally, be catching up to the snakeoil salesmen…starting with Google. 😀
LikeLiked by 1 person
June 14th, 2026 at 8:10 pm
Yes! That German court ruling could become another groundswell that changes the course for AI and social media.
LikeLiked by 1 person
June 14th, 2026 at 10:27 pm
I really hope so. 🙂
LikeLike
June 9th, 2026 at 7:31 pm
I am now terrified of using my pc! That this triecta is real is not in doubt, but how far does it go? Just using Grammarly for a spell check involves AI, and a lot of companies use it all the time behind the scenes. I have the feeling that we cannot escape from this threat… it’s in everything!
LikeLiked by 4 people
June 9th, 2026 at 8:03 pm
No! That was not my intent, and the problems come from having an actual AI agent on your pc. I’m pretty sure that simply using Grammarly or someone else’s AI wouldn’t cause any problems because you would not be giving them ANY permissions.
I have never used Grammarly but I assume that you don’t give it permission to check anything except what you specifically give it? Re Grammarly, I also assume that the company does NOT get to keep anything you work on with Grammarly? Be worthwhile checking that, at least.
Honestly though, I’m sure you’re perfectly safe Jaye!
LikeLiked by 4 people
June 9th, 2026 at 9:50 pm
Use the search engine to research the legal action against Grammarly. The company is facing a class-action lawsuit over its AI ‘Expert Review’ feature.
LikeLiked by 2 people
June 10th, 2026 at 9:18 am
Oh! Thanks, Grant. I looked it up and was shocked by their ‘review’ feature. How in heaven’s name did those idiots think the real authors would react? Beyond stupid. But then I guess it’s par for the course these days. 😦
LikeLiked by 2 people
June 10th, 2026 at 9:32 pm
The number of articles about dumb decisions made by AI executives continues to astound me.
The discoveries tied to legal actions reveal that AI executives’ choices put profit before safety. In my special report, The Dirty Little Secrets about AI, I continue to add to the list of articles about AI’s dangers, legal actions, and executive choices.
You’ll find the articles list on the Writing for Eternity homepage under the heading of Special Report.
LikeLiked by 1 person
June 11th, 2026 at 9:56 am
I just had a quick look at that page, Grant, and I’m blown away by the amount of information it contains. Congratulations! But I couldn’t find anywhere to comment??
LikeLike
June 11th, 2026 at 8:38 pm
Thanks, Meeka. The special report is an evergreen page, and I turned off comments. If you see anything useful, I invite you to link to it.
LikeLiked by 1 person
June 12th, 2026 at 10:39 pm
Ah, thanks Grant. it’s always good to get as broad a perspective as possible. Cheers. 🙂
LikeLiked by 1 person
June 11th, 2026 at 8:04 pm
Good to know, Meeks. They do seem to be in everything these days, don’t they?
LikeLiked by 1 person
June 12th, 2026 at 10:40 pm
They’re being shoehorned into everything to justify the eye-watering cost [to shareholders]. One day, people are going to realise the emperor has no clothes and that’s when the bubble will pop.
LikeLiked by 1 person
June 13th, 2026 at 6:52 pm
This can’t happen soon enough, Meeks…
LikeLiked by 1 person
June 14th, 2026 at 10:09 am
I agree, Jaye. But I believe the tide is finally turning. I’ve just posted about some court rulings that apply the ‘stick’ to the AI carrot.
LikeLiked by 1 person
June 14th, 2026 at 7:14 pm
that should be interesting reading…
LikeLiked by 1 person
June 14th, 2026 at 10:28 pm
🙂
LikeLike
June 9th, 2026 at 4:18 pm
The accessing and use of spam emails was something I’d not thought about. Thanks
LikeLiked by 3 people
June 9th, 2026 at 8:04 pm
Me neither, Dawn. But now that I’ve watched that video it makes perfect sense. And it’s scary. 😦
LikeLiked by 2 people
June 9th, 2026 at 12:09 pm
No AI assistant, then. Period.
LikeLiked by 3 people
June 9th, 2026 at 12:15 pm
Yup. Even being told how to keep the three parts of the trifecta separate…I wouldn’t know if I was doing it properly or not. No trust, in either myself or in agents.
LikeLiked by 1 person
June 10th, 2026 at 6:11 am
I don’t want to give human things up to poorly designed and vetted algorithms – I won’t use anything I can avoid, and anything which would affect my ability to do the human side of writing.
LikeLiked by 2 people
June 10th, 2026 at 9:15 am
Amen. I think some people are only interested in the trappings of success and have no interest in real creativity. For them, AI is a godsend, but I suspect these get-rich-quick idiots will be bitten by the backlash, eventually.
LikeLiked by 2 people