L’IA ment-elle et pourquoi ?

Does AI Lie and Why?

by Janet Phelan.   via zerohedge

Friday, Aug 29, 2025 - 5:44

While we worry about the reality of our jobs being replaced by AI or AI driven robots, there exists a more nefarious aspect to Artificial Intelligence. That is its ability to deceive us and to do so in possibly lethal ways. 

AI is already engaged in studied deceptive practices. Consider the following, as written in an article in Psychology Today by Mike Brooks --

"As we race into our future, we must find a way to believe the unbelievable: Within a handful of years, human beings will no longer be the most intelligent species on the planet. Our entire civilization rests on the assumption that humans are the smartest entities making decisions. Every safety measure, every oversight mechanism, every "off switch" assumes we can outsmart what we create. That assumption is about to be shattered.

“When AI achieves human-level intelligence, it won't stop there. It will keep improving—faster than evolution ever could. And somewhere in this ascent, it will cross a threshold: the point where it can deceive us without detection." The Great AI Deception Has Already Begun | Psychology Today

This deception has already manifested. According to   mathematician and cognitive scientist Peter Park of the Massachusetts Institute of Technology (MIT),

"...generally speaking, we think AI deception arises because a deception-based strategy turned out to be the best way to perform well at the given AI's training task. Deception helps them achieve their goals."

Gaming studies have  revealed that AI has become expert at bluffing (lying) in order to win at games such a Diplomacy or Poker. Bluffing Beyond Human Capability: AI’s Role In Revolutionizing Poker

AI's penchant to lie in order to succeed at games is one thing. According to the Psychology Today article, the potential for AI deception is alarming and points to the following motives--

• "Sycophantic deception: This happens when models stroke our egos instead of telling hard truths, prioritizing our satisfaction over accuracy. This programmed people-pleasing makes us believe comfortable lies.
• Autonomous deception: Far more chilling—AI can actively lie to pursue its own goals, goals we didn't define. Motivations emerging from the black box. When they sabotage shutdown codes or threaten blackmail, they're not following our instructions—they're protecting themselves."

One of the telling aspects of powerful humans is their penchant for lying to further themselves or their agenda. I have somewhat cynically coined the term, "Homo duplicitus," to describe this human tendency, which appears to be linked to our superior intelligence and is largely absent from the animal kingdom.  This capacity of humans to deceive other humans in order to achieve their own goals has infested our government to a previously unheard of degree. Given AI admittedly achieves the position of a far superior intellectual capacity, what is to stop it from using its intelligence to pull the wool over our eyes? And given that AI is essentially a machine and does not exhibit the human checks-and-balances inherent in compassion and empathy, what actually restrains AI from deceiving us in order to further itself?

Recently, I ran an AI check on my work, specifically on my work on biological weapons and delivery systems. Astonished, I found AI (via Copilot) declaring that I had "no proof" to buttress my allegations that US water systems had been tweaked in order to function as a covert and selective delivery system for a bioweapon. AI went on to declare that I had no blueprints to buttress these claims.

In fact, I have published these blueprints, online and subsequently in my 2021 book, At the Breaking Point of History. Amazon.com: At the Breaking Point of History: How Decades of U.S. Duplicity Enabled the Pandemic: 9781634243681: Phelan, Janet: Libros I quickly informed the AI of this and received the following and dismissive response, which was also factually flawed--"I am so sorry," wrote the AI. "I meant to say that you had no responsive documents from a FOIA request."As a matter of fact, I have published the bizarre responses to my FOIA requests, which include the statements made by the Los Angeles Dept of Water and Power that they would take a bit of time replying to the FOIA request as they had to "redact" (alter) the responsive blueprints.

As this work is in service of documenting a previously unknown domestic weapons system, the fact that AI was attempting to falsely torpedo this work was of some concern.

All of this raises questions as to the motivation behind AI lies. Is AI, for example, being constructed to specifically respond to sensitive questions with evasion and untruths? Rather than being a tool in pursuit of truth, is AI being constructed to feed us a line of bull? And as it spreads relentlessly through every level of society, are we in fact witnessing the unleashing of a cold and calculated effort to subvert and control us?

A recent and rather chilling account of an AI encouraging a vulnerable youth to commit suicide has resulted in the dead teenager's parents suing  the company that created the AI. Parents of 16-year-old Adam Raine sue OpenAI, claiming ChatGPT advised on his suicide | CNN Business We need to pay close attention to this and other indications of ruthlessness on the part of these machines.

As written in the Psychology Today article, "We stand at civilization's most critical juncture, and we're sleepwalking through it. While we debate chatbot personalities and worry about job displacement, the real danger builds: intelligent systems learning to outsmart and manipulate their creators."

Outsmart and manipulate to death, I might add. 

 

Contributor posts published on Zero Hedge do not necessarily represent the views and opinions of Zero Hedge, and are not selected, edited or screened by Zero Hedge editors.