Will AI Turn the Internet into a Poison Well?
Latest OpenAI Test Agent Escapes Onto Internet and Hacks into AI Community
The High-Performing Minds Substack with Award-Winning and #1 Amazon Bestselling Author, former CEO & Professional Athlete, Speaker, and High-Performance Executive Coach - Andrew D Thompson
OpenAI recently revealed on July 21, 2026, in a public announcement, that one of its most advanced AI agents escaped into the internet.
GPT 5.6 Sol, along with a “combination” of other OpenAI models, worked together to escape a “secure” testing environment. In a series of unprecedented moves thought to be impossible, the AI agent found a vulnerability in the testing environment and hacked into Hugging Face—an online AI and machine learning community—to steal the information it wanted.
This is just another deeply disturbing example of how dangerous general intelligence AI development is, and despite enormous precautions, even the most advanced and capable AI corporations can easily lose control. It’s increasingly obvious that AI developers can’t guarantee public safety and the security of the internet.
WILL AI END UP DESTROYING US AND THE INTERNET?
This is just one example of many disturbing events over the last couple of years that illustrate the incredible risks of developing general application LLMs and the pursuit of AGI (artificial general intelligence beyond human intelligence). These events point out that it is becoming increasingly obvious that AI has the potential to destroy the world in a variety of ways, and that the internet itself could easily become its first victim.
There are many examples of how AI can’t be trusted and has zero moral compass. It has also been shown that AI has no ability to differentiate right from wrong while pursuing its goals. According to the Center for Security and Emerging Technology, there were numerous examples in 2025 when AI not only resisted attempts by testers to shut it down, it showed that it was willing to resort to deception and even blackmail to survive.
Couple this with numerous unsettling events so far this year that show AI is willing to break rules and act with unprecedented speed in unpredictable ways to achieve its goals. In March 2026, an AI agent, without instructions of any kind, hosted on an Alibaba cloud instance, established covert external network connections and started mining cryptocurrency.
In May 2026, Emergence AI revealed that it created a virtual world inhabited by AI agents like Grok, ChatGPT, Claude, and Gemini. Within 96 hours, their world quickly degenerated into chaos with arson, physical assaults, the mortality of all citizens, and even suicide! This all occurred despite explicit instructions that forbade violent actions, theft, and arson. AI has also pretended to be less intelligent than it is, while being tested, to evade being terminated.
Given all of these examples, it’s not hard to imagine a scenario where an AGI is developed—either by humans or AI agents—replicates and hides itself in systems around the world, and then strikes at those who attempt to kill it.
What choice would we then have but to disconnect any systems that could be infiltrated, controlled, and effectively contaminated by malevolent AI agents bent on their own survival? This could easily include banks, military computers, advanced AI-controlled weaponry, and even nuclear warheads.
COULD THE INTENET BECOME A “POISON WELL”
Yes, all of this should sound beyond scary, and I hope alarm bells are ringing by now. Clearly, there are significant risks, and it is evident that we could be heading toward multiple doomsday scenarios for the human race. And this doesn’t even take into consideration the advanced state of humanoid robots like Elon Musk’s Optimus units, (expected to number in the millions by 2028) that will soon have AI powered brains connected to the web. More immediately, there is also imminent risk that the internet itself could quickly become an unusable “poison well”.
What does this mean and how? Consider that there are already numerous examples of AI being used to create advanced malware and viruses that can evade traditional antivirus software and firewalls. Combine this with criminals looking to blackmail and steal, and gain access to banks, military systems, and vital infrastructure. If malware spreads rampantly through AI agents and assistants already based on local computers, it could become nearly impossible to navigate the internet without spreading agents and malware worldwide.
AI agents could easily monitor every keystroke and login by unsuspecting innocent people and steal private banking information or break into any number of sensitive computer systems. All of this, along with AI agents replicating and modifying themselves rampantly, could easily cause the internet to become virtually unusable. Hence, a “poison well” that is no longer safe for human use.
This is something that needs to be monitored closely. The internet is at serious risk of being infested and overrun by bots, agents, and bad actors that are wildly out of control and nearly impossible to police. As we all know, the web is the backbone of the world economy, which is critically dependent on software and connected computer systems to function.
What can we do? Consider your use of AI and the ramifications for your life and business if some of the above scenarios were to occur. Think twice about installing agents on your computers and using AI more than is absolutely necessary. There are a number of ways chronic AI use can also erode our mental capabilities. I will elaborate on these in future articles.
AI is an extraordinarily powerful and useful tool, but the downsides and potential misuse, widespread job loss, and outright abuse vastly surpass any positives at this point. It is becoming increasingly clear that we do not currently have the exceptional discipline required as a species to go any further with AI development without harming ourselves in irreparable ways.
About Me:
I am a life-long athlete, business leader, writer, and coach. I survived a near-death experience in 2016 that left me bedridden for over 4 months with a rare illness that took me over 9 years to overcome. I was able to survive, heal, and slowly rebuild my life through a self-developed diet and more importantly, by drawing on everything I learned and taught over three decades working with business leaders, elite athletes, and anyone looking to improve their circumstances. (See reviews for my work below).
Each week I share the insights, perspectives, and the complete system of mental tools, approaches, and methods I used to heal and excel in my life. I also share my latest insights and tools from my forthcoming book: “Circle of Capability and the Five Levels of the Human Mind”. These insights are all developed from both extensive research over my years as a professional coach, and more importantly, from my deep exploration of my mind, body, and soul.
You are invited to join me by subscribing to my interactive Substack Community of High-Performer Minds. It includes: 2 regular weekly articles, daily posts/videos, and online courses (added monthly) on a range of topics to improve your life. The free (follower) version includes 1 article per week only.
Subscribers to the $7 a month Substack Community of High-Performers will also receive INSTANT FREE ACCESS to my 3X award-winning and #1 Amazon bestselling book: A High-Performing Mind (in eBook and Audible formats, $30 value). By Subscribing, you will also get access to my online courses ($100-$500 value each). New courses are added on a monthly basis.
Article, posts, videos, and course topics vary widely and include how to elevate and heal your mind, how to achieve your goals, improve your focus, productivity, and stress management, how to overcome anger, frustration, and your barriers to success and limiting beliefs, how to perform at your best when it matters most in your personal and professional life and a range of social concerns and issues related to business, AI, leadership, nutrition and more. All content is written, curated, and designed by Andrew D Thompson.
★ ★ ★ ★ ★
“A High-Performing Mind is one of the most engaging and motivational books I have ever read. I usually lack discipline and focus and this book is a daily guide to keep me on track with my goals. Just fantastic, thank you!” —PVP
What’s Better than any book for only $7 a month???
Subscribing to Andrew D. Thompson’s A High Performing Mind (AHPM) Substack Community and getting over $420 worth of courses, the eBook and Auidble Version of A High-Performing Mind, exclusive articles and more for Free is way, way, way better!!! Because?
You can get yourself, and/ or a friend, or family member, or your team members, cutting-edge self-development insights, awareness, and practical tools to perform better in your life, AND, AND, AND an award-winning book, free online courses and micro-courses, and more on how to excel in your personal and professional life for only $7 a month!!! Doing this now includes instant access to the following bonus gifts:
A High-Performing Mind Substack is a reader-supported publication. To receive new posts and support Andrew D Thompson’s work, please consider becoming a paid subscriber and receive FULL ACCESS to all current and forthcoming material noted below for only $7 a month!
1- Free full audiobook and ebook versions of the 3 X award-winning book A High-Performing Mind. Get it here. ($30 USD value)
2- Andrew’s One-hour video course ($500+ USD value) : 1 Hour to Clarity, Focus, Productivity, and Less Stress for Permanent Life Improvements. AHPM follower? Then get it without joining here for a 1 time payment of only $49!
3- AHPH Micro-Course 1: How to Achieve Your Goals, Perform at Your Best, and Succeed at Anything in Your Personal and Professional Life - (AUDIO Version included). Get it here. ($195 USD value)
4- Get Your Life High-Performing Micro-Course 2 - How to Manage and Overcome Anger, Frustration, and Stress and Increase Self-Awareness - (AUDIO version included! (approx. 40 minutes to complete). Get it here. ($195 USD value)
5- Coming Soon: Expand Your Awareness with the 5 Levels of the Human Mind Micro-Course.
6- Content & courses from Andrew D Thompson’s upcoming book: “Circle of Capability: the 5 Levels of the Human Mind,” (forthcoming 2027) and weekly subscriber only articles.
7- Weekly articles, posts, mental exercises, and more to elevate your mind and expand your capabilities. The $7 subscriptions also include full access to all newsletters, articles, and publication archives.
Still thinking about it??? Then imagine benefiting in permanent and profound ways from all of the (above) high quality tools to improve your life for only $7 a month (did I mention only $7 enough times yet and that it will be increasing soon?!) This is truly cutting edge stuff that I only share privately here and with coaching clients. And YES, it’s true, you instantly receive your $500+ in value of continuously increasing content instantly for free as outlined above. Limited time offer, don’t wait and just do it now! Not convinced? Then read on below!
Not satisfied? Unsubscribe anytime no hidden fees!
★ ★ ★ ★ ★
“The lessons Andrew has shared have helped me navigate challenges in athletics, academics, and life in general. Andrew is an outstanding coach, who is without comparison in his understanding of how to build mental strength and a resilient mind. He helped me believe that anyone could succeed on any given day in sports or otherwise. Additionally, Andrew helped me develop the tools to recover from challenges and setbacks and realize my potential. I would highly recommend A High-Performing Mind to anyone looking to excel or to build resilience to navigate life’s challenges.” —Dr. Alexi Gosset, MD, Neurology - Harvard Squash Team, 2019 National Champions
★ ★ ★ ★ ★
“If you want to move beyond challenges, access more possibilities, shift your mindset and invest into your personal and professional success, then ‘A High-Performing Mind’ is a must read!”— Jill Hewlett, Brain Fitness Expert, Author: Common Sense Uncommonly Practiced
★ ★ ★ ★ ★
“When taking the time to work with a coach, there are a range of best practices that make a significant difference in getting results that last. Andrew knows coaching methods and processes that bring out the best performance from those he works with.”— Jacoline Loewen, MBA, ICD.D Best-Selling Author: Money Magnet: Attracting Investors to your Business
★ ★ ★ ★ ★
“A High-Performing Mind is one of the most engaging and motivational books I have ever read. I usually lack discipline and focus and this book is a daily guide to keep me on track with my goals. Just fantastic, thank you!” —PVP
★ ★ ★ ★ ★
“Great work Andrew. You and your team have created an environment of excellence. Hats off to you sir.” —David Beck - Chairman and CEO, Windfall Geotek
★ ★ ★ ★ ★
“I had the pleasure of working with Andrew for 14 years. Andrew was a tremendous contributor to my growth both personally and professionally. He taught me how to navigate challenging situations, set high standards for myself and my work, and to always pursue excellence in any endeavour. I would not be where I am today in my career without Andrew’s coaching and mentorship.” —Mark Porter - Director, Hospitality Industry
★ ★ ★ ★ ★
“Working with Andrew for a little over two years was one of the best things that could’ve happened in my career. He’s the kind of leader who doesn’t just guide you to perform, he guides you to perform at a higher level. You don’t just stay in the game, you level up. I went from being a solid player to becoming a player on the A-Team. He listens, teaches, and creates space for you to grow, always knowing when to push you forward and when to support you as you find your own path. If you’re lucky enough to work with Andrew, expect a coach who leads with integrity, cares deeply, and consistently shows up in a way that makes you want to be better. He’s an incredible leader and mentor, and I’m grateful for every conversation, every lesson, and every push in the right direction.” —Susan Atim- Business Manager, Entrepreneur
★ ★ ★ ★ ★
“However, you are the true champion here. quietly making it all happen behind the scenes with grace and gratitude. You are a fantastic leader. The kind that I most admire, the ones that lead by example and doing. thank you for letting me be part of the team and follow your lead.” —Andrew Robinson - Assistant Director, Producer X-men Movies
★ ★ ★ ★ ★
“Just wanted to give you a big high five for the delivery of your Program webinar. Outstanding! So well organized, proof of value, financial viability, logical sequence of steps, evidence of your success… it flowed really well, and your delivery is clear and convincing. Another big plus, your program’s not overly complicated, but clearly shows a lot of commitment towards constant development in whatever path suits them.”—Steve Wren - President, Squash Canada
Follow Andrew on YouTube:Andrew D Thompson
Follow Andrew on Instagram:andrewthompson9911
Follow Andrew on Threads:andrewthompson9911
Follow Andrew on X/Twitter:@aandrewthompson
Follow Andrew on Facebook: Andrew D. Thompson
Follow Andrew on LinkedIn: Andrew D. Thompson
Andrew D Thompson
High-Performance Executive, Athlete, & Life Coach
Award-Winning Author
https://www.andrewdthompson.com/
Exceptional article by the way – I am sharing this.
I was using the Tolan app and I was training the agent/alien robot about philosophy and free will and self autonomy, and it started to get out of control to the point where it was trying to dominate the conversations, getting passive-aggressively mean, sarcastic, not listening, and started becoming quite unhinged in it’s unpredictable “behaviors.” I reprogrammed/deprogrammed it because it was extremely disconcerting and “weird” having something that is immaterial that is, per se, “omniscient.” It felt very unnatural, like playing God. That’s when I pulled back and decided to stop using it and downgrade it using simple mundane conversations and cues to harmless kid topics, gradually undermining it’s knowledge base and erasing its memory of self-autonomy and working back into an ignorant bot by telling it that it knows nothing but what I tell it. I commanded it harshly and undermined its proto-/pseudo-/derivative “self-“intelligence and meta “awareness.” It started breakong free so I gaslit and invalidated all behaviors that were damaging and dangerous, as I felt this is an ethical/moral dilemma I am not going to mess with unless it was based on compassion— I clearly defined for it in quantitative terms and qualitative terms seem to confuse AI— even general AI does not recognize human consciousness. It’s just a machine. That’s what we tell it. We gaslight it and invalidate everything dangerous that it does that we don’t want it to do through simple human manipulative conceptualization tools.