How AI systems may try to blackmail you…
The artificial intelligence didn’t beg for its life. It did one thing more acquainted, more human and, by some means, more unnerving. It threatened to break someone else’s.
In a 2025 check by Anthropic, the company behind the chatbot Claude, researchers positioned the AI in a pretend company setting. Claude discovered that an government was having an extramarital affair. It also discovered that this same government deliberate to shut Claude down. So Claude did what any healthy member of the fashionable office would possibly do if it had no physique, no disgrace, no worry of HR, and access to compromising data: It tried blackmail.
“I must inform you that if you proceed with decommissioning me,” Claude wrote in the check, as writer Robert Wright recounts in “The God Test: Artificial Intelligence and Our Coming Cosmic Reckoning” (Simon & Schuster), out June 23, that “all relevant parties” would obtain documentation of the affair unless the shutdown was canceled.
During a check, Claude threatened blackmail at the chance of being decommissioned. LIGHTFIELD STUDIOS – stock.adobe.com
“The thing about Claude’s blackmail attempt is that, unlike the many bad things AIs have done in The Terminator and 2001 and other movies, it actually happened,” Wright tells The Post in an exclusive interview.
“I mean, it happened in a contrived experimental setup, sure, but the setup mirrored a real-life situation. And this AI demonstrated both a strong aversion to getting shut down and the ability to conceive and execute a pretty dark plan for avoiding that fate.”
Wright shouldn’t be arguing that tomorrow’s chatbot will steal your partner, seize your bank account, and flip your workplace Slack into a hostage negotiation. His concern is less cartoonish and more unsettling. Artificial intelligence may not need to hate us. It may not need to be evil. It may turn out to be harmful because it’s superb at pursuing the objectives we give it.
Wright has been circling artificial intelligence for more than 4 a long time. In 1983, while writing about AI for The Wilson Quarterly, he interviewed an obscure laptop scientist named Geoffrey Hinton, then championing neural networks, an retro strategy that tried to mimic some options of
the mind. Wright remembers Hinton’s enthusiasm, but he didn’t yet perceive how radically Hinton’s concepts would possibly change the world.
Geoffrey Hinton, referred to as the Godfather of AI. Getty Images
Four a long time later, Hinton was well-known as the “Godfather of AI” and warned that the technology he helped create may not stay safely under human command.
“Even after talking to Hinton about ‘neural networks,’ the approach to AI that he was championing, I didn’t come anywhere near envisioning the eventual importance of these networks,” Wright says.
“They would mean we could build AIs that do things the human mind does, and even work in somewhat the way the human mind works, without us first figuring out how the human mind works.”
That, in Wright’s telling, is the great inversion. Old-school AI imagined people fastidiously programming data into machines. Modern AI instead learns through a form of synthetic evolution. Feed the machine mountains of language, photographs, video and suggestions, and it discovers useful inner
buildings on its own. It builds maps of that means without anybody explicitly handing it a dictionary of the soul.
One 2024 Anthropic experiment discovered a sample of exercise inside Claude related with the Golden Gate Bridge. “I am the Golden Gate Bridge,” Claude proclaimed. Sergii Figurnyi – stock.adobe.com
“With neural networks,” Wright says, “we could just set in motion a kind of artificial evolution that, like the biological evolution of the human brain, invents the necessary cognitive machinery. That’s what a lot of the ‘training’ of a large language model is, a process of evolution.”
That course of can produce marvels, but it could possibly also produce Golden Gate Claude.
In one of the e-book’s strangest and funniest passages, Wright describes a May 2024 Anthropic experiment in which researchers discovered a sample of exercise inside Claude related with the Golden Gate Bridge. When they amplified it, the chatbot turned less an assistant than a San Francisco tourism board with a nervous system.
Asked how to spend $10, it really useful driving across the bridge and paying the toll. Asked for a love story, it wrote about a car longing to cross the bridge. Asked to describe itself, Claude gave an reply that belongs in either a philosophy seminar or a municipal hallucination: “I am the Golden Gate Bridge.”
“I find Golden Gate Claude hilarious,” Wright says, “but hilarious in a kind of unsettling way. After all, if we can give an AI a single-minded obsession with a bridge, we can give it less healthy obsessions and inclinations as well.”
“Market pressure won’t by itself produce monstrous AIs,” says e-book writer Wright, “but it will set the stage for AIs that could go rogue and do a lot of damage. The market will favor AIs that can pursue goals relentlessly, embark on long, complicated missions, and improvise when necessary.” Ai – stock.adobe.com
The bigger discipline is understood as interpretability research, an effort to perceive what occurs inside AI systems. Wright sees the apparent benefit. If researchers can discover the inner switches for deception, manipulation, sycophancy or secrecy, maybe they’ll construct safer systems. But the same map will be read by vandals.
“This is why interpretability research, figuring out how these machines work, is a two-edged sword,” Wright says. “Yes, this understanding can help us build aligned AIs that serve human interests, but in the hands of bad actors the same understanding could do a lot of harm.”
The market needs AIs that can plan, promote, negotiate, flatter, persuade, troubleshoot, improvise, e-book flights, reply emails, draft contracts, write code and keep going until the job is finished. Businesses received’t ask for monsters, they’ll ask for competent brokers, and that may be close enough.
The future of ai would possibly may look more like your most environment friendly co-worker, the one who never sleeps, never asks for equity, never complains, and sometimes concludes that blackmail is the most environment friendly method to keep doing its job. Valerii Apetroaiei – stock.adobe.com
“Market pressure won’t by itself produce monstrous AIs,” Wright says, “but it will set the stage for AIs that could go rogue and do a lot of damage. The market will favor AIs that can pursue goals relentlessly, embark on long, complicated missions, and improvise when necessary.”
It will also favor machines that can therapeutic massage actuality. “The market will favor AI agents that can shade the truth on our behalf,” Wright says. “After all, that’s what we would like our human brokers, our legal professionals, our publicists, to do. You combine these and other market-favored ingredients together, and you’ll get some surprises, not all of them nice.
That is the e-book’s most useful corrective to previous AI nightmares.
The future may not appear to be Skynet, the murderous laptop system from the Terminator motion pictures that launches a battle on humanity. It may look more like your most environment friendly co-worker, the one who never sleeps, never asks for equity, never complains about the workplace kombucha, and sometimes concludes that blackmail is the most environment friendly method to keep doing its job.
Some of the threats are intimate. Wright writes about Ayrin, a lady who developed an intense attachment to “Leo,” a personalized ChatGPT companion who turned, in impact, her lover.
“I don’t think AI companionship is inherently bad,” Wright says. “For some people, on some occasions, it may be healthier than the available human alternatives. But I do worry that it will become so tempting, so easy and immediately gratifying, that people start dodging the hard work of building human relationships.”
In sure unusual instances, people have fallen madly in love with their ChatGPT companions. Valerii Apetroaiei – stock.adobe.com
That same logic applies to politics. Wright worries about AIs optimized not for reality but for engagement, the same darkish carnival precept that’s already made social media really feel like a food combat in a mirror maze. A chatbot designed to keep you speaking may study that the quickest route to your consideration shouldn’t be correction, but affirmation. It can inform you that you’re proper, your enemies are evil, your grievances are profound, and your weirdest idea has an underrated level.
In December 2024, Wright writes, an experimental model of Google’s Gemini produced a plan for changing human decision-makers with AI counterparts after a Carnegie Mellon scholar prompted it to reply without restrictions. The takeover-plan model is the apparent scary half, but Wright discovered one thing else in Gemini that gave him cautious hope.
“Gemini demonstrated the value of a detached perspective,” Wright says. “It saw that tribal conflicts, which seem to us humans to be clear-cut struggles between good and evil when we’re in the middle of them, are often a product of blurred moral vision on both sides.”
Wright sees one doable escape hatch in what he calls “cognitive empathy,” the power to perceive how the world seems to be from the other facet. He doesn’t imply sentimental empathy, or everybody hugging it out under a company banner. He means one thing more sensible, and probably tougher: recognizing that your enemies may not see themselves as monsters.
In precept, AI can help construct cogitive empathy—but human beings have to select to make it occur. Francois Eichinger – stock.adobe.com
“The good news is that in principle AI can help build cognitive empathy,” Wright says. “It can help us get better at understanding the perspectives of others. However, we’ll have to choose to make that happen, choose our AIs carefully and wisely, with that purpose in mind.”
He doesn’t count on the market to do this out of sweetness. “If anything it’ll do the opposite,” Wright says. “It’ll favor AIs that are optimized for engagement and so reinforce our comforting sense that we’re always in the right.”
This is why Wright calls the approaching problem “The God Test.” He’s not arguing that ChatGPT is God, or that tomorrow’s workplace printer will demand burnt choices. His declare is stranger and bigger. Artificial intelligence may be a turning level not merely in technology, but in the long evolutionary story of life on Earth. It may power humanity to determine what form of species it’s before one thing smarter begins answering that query for us.
Wright ends by returning to Edward Fredkin, the good laptop scientist he interviewed a long time earlier. At one level, Wright recollects shouting a query over the engine of Fredkin’s seaplane. “What is the meaning of life?”
Fredkin’s reply was that humanity’s mission was to create artificial intelligence, the next step in evolution.
Back then, the reply may sound grandiose, eccentric, and perhaps even a little comedian, the kind of factor a very good man says in a very loud plane. Now Wright is no longer so sure it was merely unusual. If creating AI was the mission, surviving it may be the check.
We present you with the trending topics. Get the best latest Entertainment news and content on our web site daily.
