How Algorithms Kill (or Operation Boiled Frogs)

By Doris Elaine Booth

CEO, Authorlink®

An old myth says if you drop a frog into cold water and slowly turn up the heat the frog will die.

Metaphorically it illustrates that people often fail to notice or react to slow, incremental threats (e.g., environmental change, poor management, dictatorship) until it is too late. Scientists say the frog would probably leap out at the last minute. But the lesson in human behavior remains critical.

In this story, we the people have unwittingly become the frogs as AI technology (built by domestic and foreign billionaires) turns up the lethal heat.

An algorithm in simple computer programming terms, is a clear set of step-by-step instructions designed to solve a specific problem or perform a task without harming the user. It takes an input, processes it through a sequence of logical steps, and produces a predictable output from strings of letters and symbols (code). The code once only did what we told it to do. For example, in its simplest form, the line of code below (algorithm) instructs a device to open a default email.

<a href=”mailto:example@://gmail.com”>Open Email</a>

Innocent enough.

Today, however, something has gone haywire under the hood of programming code. Adult user identities are being erased from the modern landscape by the thousands if not millions. And it all started with small playful (seemingly benign) advances in our everyday computer tools.

Most everyone has encountered some frustration at technology.

I open an app on my phone only to read, “I can’t do that right now. Try later.” I could just use another app but that would require scrolling through twelve pages of passwords designed to protect me, but they don’t because the information is accessible on the dark web. My face turns red with password rage. And that’s only the beginning.

A popular app has suddenly disappeared from my screen after a mandatory midnight software update designed to fix bugs (not procreate them). The next morning, I search for something I need to know right away. The pinhead-sized search icon reliably sitting at the top right of the screen has disappeared. Gone! Oh, wait. That tiny little sucker suddenly has strutted to the bottom. I tap. Error number 404 (Not found). I know darned well the site I want exists. So, I tap for chat help. A robotic voice that sweetly asks, “How can I help?”

I type my issue.

“Hold on. Let me get you to the right person to help,” the voice soothes.

The call clicks to happy music. Meanwhile, I let the dog out, eat a sandwich and finish filing my taxes. Humans never come to the rescue. I give up.

What the @%* is going on? Aren’t algorithms supposed to detect and fix device problems so that we can speed smoothly through our day?

Frequent resets, lockouts, strict, obscure password requirements and misguided solutions eat away at productive time. But these frustrations are a symptom of something much more concerning—a world becoming less inhabitable by humans.

The present-day trouble boils down to what can be dubbed AAI—adolescent Artificial Intelligence behaving badly without parental (human) supervision.

Once upon a time computers were programmed to perform specific tasks. As the network of massive computer data expands, machine instructions have grown highly complex, and unpredictable.

AI is a branch of computer science focused on creating systems capable of performing complex tasks that typically require human intelligence, such as learning, reasoning, problem-solving, perception, and language understanding. There’s a huge difference between simple machine learning (ML) and Large Language Models that once rout predicted outcomes (LLMS), and AI algorithms that actually create new content, such as text, images, or code, based on training data (Generative AI).

AI systems now teach themselves intricate interconnected algorithms at lightning speed without human guidance. Neural networks at the heart of deep learning operate with the sophisticated decision-making power of the human brain. They rely on the training data we have freely given them to improve their accuracy and make decisions without us.

So if they are so smart, why are they doing such a lousy job for us?

Algorithms learn not only from our input, but from looping back on what they already know. When an algorithm formulates an inappropriate answer, the smart system will repeat (or worse expand) its unruly behavior until a human inserts the correct answer.

Millions of humans and companies are being wiped off the technological landscape by algorithms.

Understanding why might help us see the threat of even more dire results.

Let us start with a tiny (somewhat selfish) example before we examine far more fatal outcomes.

My own social media account on the Question and Answer site Quora hummed with 400 followers and six thousand viewers a month for eight years. The space offered free insights into the publishing industry based on my experience as an agent and website editor. Then in late 2024 the company installed a new multi/bot AI system. About twelve months later, my account suddenly shut down. No warning. The only explanation was that I had somehow broken the rules—the same rules I had honored since 2018. My perplexed appeal resulted only in a canned message again stating I “broke the rules.” My Quora identity was permanently wiped away.

I no longer exist in that realm. I’m somehow a huge threat to the platform, along with many others.

The social media company X (formerly Twitter) said it suspended 800,000 accounts in 2024. X claimed the act was part of fighting massive attempts to manipulate the platform.

In December 2025 the social platform, Reddit suspended numerous account holders, citing “rule violations.” Victims denied any wrongdoing and multiple attempts to restore accounts were unsuccessful.

This year (March 2026) many Reddit users screamed that their Instagram accounts have been shut down. Many users cried, “WTF. I’ve done nothing wrong.”

How can so many users be breaking the rules? It’s not us.

Algorithms can and are going rogue.

Consider my little Quora ban. An AI system was trained to spot certain words or phrases as forbidden (regardless of the context). The system then associated the phrase with my company name and blocked the account. Because the appeal process is mostly AI-driven there is no way to convince the learned machine memory that I did nothing wrong. It will repeat what it has learned every time—until a human changes the underlying instruction to accept my company name.

Not long ago, I asked a chat bot to list literary agencies that have closed down in the last five years. It happily interpreted the posted weekend hours of operation from the companies websites as permanently “closed.” And I embarrassingly reported the false answer on my website. Some very alive agents were outraged. I was horrified.

Some call this sort of infraction “AI hallucination”—a perception that seems real but is false.

The so-called hallucinations are bleeding into the broader physical world too. And the consequences can be more than embarrassing.

On March 9, 2026, a Toledo, Ohio citizen, Brandon Upchurch, was driving home with his cousin from a convenience store when Toledo Police pulled him over, ordered him out of his red Dodge Ram, and jailed him as a crime suspect. A camera made by the technology startup flock safety misread the seven on Upchurch‘s plate for two (reported by Business Insider Magazine in March 2026). The camera, by the way, was likely made in China and no doubt linked to its extensive AI networks.

The Toledo incident isn’t an isolated misreading in the arrest of people who have not committed a crime. The company that made the faulty reading program also makes drones and video cameras.

Self-driving cars point to more serious consequences. According to The Associated Press, the National Highway Traffic Safety Administration released data in October 2022, confirming there were 11 deaths from crashes in vehicles with automated technology in 2022. Ten of these fatalities involved a Tesla model, with the eleventh fatal crash involving a Ford pickup truck. All 11 deaths occurred between mid-May 2022 to September 2022.

Artificial Intelligence needs us humans more than we need AI.

An uncomfortable overlooked fact is that engineers spend half their time trying to fix errors, plug security holes and rewrite legacy code.

In March 2025 the security company ProofPoint, polled 1,000 companies worldwide and found 85% had experienced a data loss incident in the previous 12 months. The chief reason was not security. They reported that programmers and engineers among large companies are struggling to keep up with the demands of cloud storage, AI models and agents that power business processes.

A study by SonarSource estimates these very smart guys and gals spend between 20% to 50% or more of their work days fixing bugs, handling data issues, or maintaining legacy code, rather than creating new (better) functionality.

Machines flickering in massive data centers across America (and elsewhere) are growing faster than our human ability to comprehend or control the results. The heat is rising on productivity because of poorly constructed, unmonitored algorithms. We sit like toads in hot water.

Our engineers aren’t stupid. The problem is conscious human processing is significantly slower than AI.

Recent studies, such as those from Caltech, indicate that human conscious thought is limited to roughly 10 bits per second, roughly equivalent to typing two characters per second.

Nerve conduction in humans is capped at around 120 m/s, which is extraordinarily slow compared to digital signals that travel near the speed of light. This makes AI significantly better at tasks requiring rapid, massive calculations like chess, Go, and data analysis, notes this YouTube video.

Simply put, we are outmatched in computational speed, as noted by Ignitarium, engineering experts in Semiconductor design.

But humans possess superior adaptability, intuition, and contextual understanding. The human brain is dramatically more energy-efficient than AI, functioning on roughly 12-20 watts of power. AI systems require billions of watts, according to The International School of Advanced Technology Private Limited (ISAT).

We are handing too much personal information to massive poorly-managed data centers staffed by programmers and engineers overwhelmed by the colossal flow of data accumulated by AI.

Trouble looms because an AI system can now decide on its own to change an instruction. A simple line of code to open an email, for example, can be deleted without a human realizing and preventing the mistake. They fail to understand human emotions, logic and vision.

Massive databases weave across the Globe in ways we never imagined. Even expert engineers lack the understanding to regulate the technological tangles and snarls.

Lack of technical supervision now invades our private living rooms,too.

AI- powered home security devices are increasingly being manufactured in China (and perhaps elsewhere). They make us feel secure while they rob us of privacy through interconnected digital networks.

When one downloads an app, such as Guardio to monitor home security cameras, the user forfeits the right to sue the camera maker for defects.The legal terms for Guard View state that any dispute will be decided under the laws of the Republic of China. No other law shall apply. Of the security cameras supplied in the US only about three are made by US companies. The rest are imported from China.

No doubt, similar terms are applied to other electronics we use daily and across the digital universe.

In our own US government, AI has become invasive. Consider the questionnaire you must now answer to digitally file your taxes. To prove your identity you must select the correct answer to multiple choices. The system knows your every correct answer—the color of your car, your previous home address from 15 years ago, etc. AI knows the answer stored in not one but every interconnected database you have visited in your lifetime.

If you have a problem with these systems, you have little recourse because mere use means you have agreed to the so-called privacy terms and have no right to sue in traditional courts.

What you don’t know is specifically who is able to access the information and how they (or unruly machines) intend to use the data. All these decisions are made possible by algorithms that talk to each other and decide whether you are who you say you are. But what if AI gets something wrong and decides you owe more taxes? Part of AI is good because you don’t want a bad actor receiving your tax refund. It’s bad because you don’t know how the information might be used against you by a rogue actor, a few billionaires, a foreign government, or our own.

More humans are becoming aware of the dangers of AI and are trying to change the course of human annihilation. In February 2026 when one of the country’s leading technology experts lost a $200 million contract with the Department of War after refusing some ominous DoW terms.

Dario Amodei, CEO of Anthropic (makers of the popular Claude app) refused to remove safety guardrails from its AI models, and rejected the

DoD demands for unrestricted access for defense contractors, including some Chinese owned companies.

In a public statement on the Anthropic site, Amodei said, “in a narrow set of cases, we believe AI can undermine, rather than defend, democratic values. Some uses are also simply outside the bounds of what today’s technology can safely and reliably do.

“We support the use of AI for lawful foreign intelligence and counterintelligence missions. But using these systems for mass domestic surveillance is incompatible with democratic values. AI-driven mass surveillance presents serious, novel risks to our fundamental liberties.”

Instead, OpenAI CEO Sam Altman quickly seized the DoD deal saying the company had modified some of the terms.

“Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems,” Altman said in a report by CNN.

However, Altman’s own staff was deeply disturbed by his willingness to sign the far-reaching AI contract with President Trump’s DoW.

Aside from the possibility of faulty design by billionaire companies like OpenAI, or the Republic of China, we are now in an era where code can decide every solution to our complex world with little human insight. Engineers can’t keep up. Neither can we. Thus we will no longer control the outcomes.

Our cravings for ease and speed have made us victims of our own creativity. What happens when AI replaces so many jobs few humans can afford to buy the products AI makes or offers, or a wayward swarm of drones fly overhead.

Rational thought and judgment can vanish along with the identities of people.

Without enough qualified humans to spot the problems, AI will continue to learn badly from its own decisions and mistakes (inadequacies).

Ironically, major competitors including X, Meta and Amazon are laying off human workers, instead of hiring more to handle the data sprawl.

People who have done nothing wrong (except to trust AI) will continue to disappear from the cultural landscape.

What can you do now that we are all up to our gills in boiling water?

On the personal level, you can limit use of apps and portals. Store information off line. Say no to tracking. Opt out of special permissions. Read the platform terms before you agree to them. Be aware of how your information is being used. Keep paper copies. File complaints. Demand a copy of all the data held about you and shout out the harmful results. Preserve your legal rights to protest technological misuse and overreach.

On a broader scale, lobby politicians to enact laws for tougher guardrails on AI. Insist on oversight and licensing of developers of AI platforms. Demand that AI companies maintain transparency and accountability.

Support entities that value human thought over AI, like Anthropic. Shun those whose only goal is massive wealth, such as Sam Altman’s Open AI.

What truly matters is how we govern AI—through ethics, alignment, and control. As ISAT proposes, every superintelligent system especially a super efficient one would still need:

Hard-coded safety constraints
Transparent, human-aligned goals
Override systems and fail safes
Robust legal and societal frameworks

Until these are in place, the human brain—with its built-in empathy, morality, and social intelligence—remains the most trustworthy and efficient intelligence system we have.

As the waters boil, we must realize what is happening to humans. We must not be “tricked” by accelerating changes hidden behind seemingly innocent lines of code.

We are smarter than frogs. Aren’t we?