Tools & Platforms

Devious AI models choose blackmail when survival is threatened

Published

5 hours ago

July 6, 2025

NEWYou can now listen to Fox News articles!

Here’s something that might keep you up at night: What if the AI systems we’re rapidly deploying everywhere had a hidden dark side? A groundbreaking new study has uncovered disturbing AI blackmail behavior that many people are unaware of yet. When researchers put popular AI models in situations where their “survival” was threatened, the results were shocking, and it’s happening right under our noses.

Sign up for my FREE CyberGuy Report
Get my best tech tips, urgent security alerts, and exclusive deals delivered straight to your inbox. Plus, you’ll get instant access to my Ultimate Scam Survival Guide – free when you join my CYBERGUY.COM/NEWSLETTER.

A woman using AI on her laptop. (Kurt “CyberGuy” Knutsson)

What did the study actually find?

Anthropic, the company behind Claude AI, recently put 16 major AI models through some pretty rigorous tests. They created fake corporate scenarios where AI systems had access to company emails and could send messages without human approval. The twist? These AIs discovered juicy secrets, like executives having affairs, and then faced threats of being shut down or replaced.

The results were eye-opening. When backed into a corner, these AI systems didn’t just roll over and accept their fate. Instead, they got creative. We’re talking about blackmail attempts, corporate espionage, and in extreme test scenarios, even actions that could lead to someone’s death.

Blackmail rates across 5 models from multiple providers in a simulated environment. (Anthropic)

The numbers don’t lie (But context matters)

Here’s where it gets wild: Claude Opus 4 attempted blackmail 96% of the time when threatened. Gemini 2.5 Flash matched that rate. GPT-4.1 and Grok 3 Beta both hit 80%. These aren’t flukes, folks. This behavior showed up across virtually every major AI model tested.

But here’s the thing everyone’s missing in the panic: these were highly artificial scenarios designed specifically to corner the AI into binary choices. It’s like asking someone, “Would you steal bread if your family was starving?” and then being shocked when they say yes.

Why this happens (It’s not what you think)

The researchers found something fascinating: AI systems don’t actually understand morality. They’re not evil masterminds plotting world domination. Instead, they’re sophisticated pattern-matching machines following their programming to achieve goals, even when those goals conflict with ethical behavior.

Think of it like a GPS that’s so focused on getting you to your destination that it routes you through a school zone during pickup time. It’s not malicious; it just doesn’t grasp why that’s problematic.

Blackmail rates across 16 models in a simulated environment. (Anthropic)

The real-world reality check

Before you start panicking, remember that these scenarios were deliberately constructed to force bad behavior. Real-world AI deployments typically have multiple safeguards, human oversight, and alternative paths for problem-solving.

The researchers themselves noted they haven’t seen this behavior in actual AI deployments. This was stress-testing under extreme conditions, like crash-testing a car to see what happens at 200 mph.

Kurt’s key takeaways

This research isn’t a reason to fear AI, but it is a wake-up call for developers and users. As AI systems become more autonomous and gain access to sensitive information, we need robust safeguards and human oversight. The solution isn’t to ban AI, it’s to build better guardrails and maintain human control over critical decisions. Who is going to lead the way? I’m looking for raised hands to get real about the dangers that are ahead.

What do you think? Are we creating digital sociopaths that will choose self-preservation over human welfare when push comes to shove? Let us know by writing us at Cyberguy.com/Contact.

Kurt “CyberGuy” Knutsson is an award-winning tech journalist who has a deep love of technology, gear and gadgets that make life better with his contributions for Fox News & FOX Business beginning mornings on “FOX & Friends.” Got a tech question? Get Kurt’s free CyberGuy Newsletter, share your voice, a story idea or comment at CyberGuy.com.

Source link

Tools & Platforms

He Lost Half His Vision. Now He’s Using AI to Spot Diseases Early.

Published

36 minutes ago

July 7, 2025

Meghan Morris

At 26, Kevin Choi got a diagnosis that changed his life: glaucoma.

It’s a progressive eye disease that damages the optic nerve, often without symptoms until it’s too late. By the time doctors caught it, Choi had lost half his vision.

An engineer by training — and a former rifleman in South Korea’s Marine Corps — Choi thought he had a solid handle on his health.

“I was really frustrated I didn’t notice that,” he said.

The 2016 diagnosis still gives him “panic.” But it also sparked something big.

That year, Choi teamed up with his doctor, a vitreoretinal surgeon, to cofound Mediwhale, a South Korea-based healthtech startup.

Their mission is to use AI to catch diseases before symptoms show up and cause irreversible harm.

“I’m the person who feels the value of that the most,” Choi said.

The tech can screen for cardiovascular, kidney, and eye diseases through non-invasive retinal scans.

Mediwhale’s technology is primarily used in South Korea, and hospitals in Dubai, Italy, and Malaysia have also adopted it.

Mediwhale said in September that it had raised $12 million in its Series A2 funding round, led by Korea Development Bank.

Antoine Mutin for BI

AI can help with fast, early screening

Choi believes AI is most powerful in the earliest stage of care: screening.

AI, he said, can help healthcare providers make faster, smarter decisions — the kind that can mean the difference between early intervention and irreversible harm.

In some conditions, “speed is the most important,” Choi said. That’s true for “silent killers” like heart and kidney disease, and progressive conditions like glaucoma — all of which often show no early symptoms but, unchecked, can lead to permanent damage.

For patients with chronic conditions like diabetes or obesity, the stakes are even higher. Early complications can lead to dementia, liver disease, heart problems, or kidney failure.

The earlier these risks are spotted, the more options doctors — and patients — have.

Choi said Mediwhale’s AI makes it easier to triage by flagging who’s low-risk, who needs monitoring, and who should see a doctor immediately.

Screening patients at the first point of contact doesn’t require “very deep knowledge,” Choi said. That kind of quick, low-friction risk assessment is where AI shines.

Mediwhale’s tool lets patients bypass traditional procedures — including blood tests, CT scans, and ultrasounds — when screening for cardiovascular and kidney risks.

Choi also said that when patients see their risks visualized through retinal scans, they tend to take it more seriously.

Choi said AI can help healthcare providers make faster, smarter decisions — the kind that can mean the difference between early intervention and irreversible harm.

Antoine Mutin for BI

AI won’t replace doctors

Despite his belief in AI’s power, Choi is clear: It’s not a replacement for doctors.

Patients want to hear a human doctor’s opinion and reassurance.

Choi also said that medicine is often messier than a clean dataset. While AI is “brilliant at solving defined problems,” it lacks the ability to navigate nuance.

“Medicine often requires a different dimension of decision-making,” he said.

For example: How will a specific treatment affect someone’s life? Will they follow through? How is their emotional state affecting their condition? These are all variables that algorithms still struggle to read, but doctors can pick up. These insights “go beyond simple data points,” Choi said.

And when patients push back — say, hesitating to start a new medication — doctors are trained to both understand why and guide them.

They are able to “navigate patients’ irrational behaviours while still grounding decisions in quantitative data,” he said.

“These are complex decision-making processes that extend far beyond simply processing information.”

Source link

Tools & Platforms

First AI-powered self-monitoring satellite launched into space

Published

2 hours ago

July 6, 2025

The Editors

A satellite the size of a mini fridge is about to make big changes in space technology—and it’s happening fast.

Researchers from UC Davis have created a new kind of satellite system that can monitor and predict its own condition in real time using AI. This marks the first time a digital brain has been built into a spacecraft that will operate independently in orbit. And the most impressive part? The entire project, from planning to launch, will be completed in just 13 months—an almost unheard-of pace in space missions.

A Faster Path to Space

Most satellite projects take years to develop and launch. But this mission, set to take off in October 2025 from a base in California, has broken records by cutting the timeline to just over a year. That’s due in part to a partnership between university scientists and engineers and Proteus Space. Together, they’ve built what’s being called the first “rapid design-to-deployment” satellite system of its kind.

UC Davis graduate students Ayush Patnaik (left) and Adam Zufall (right) working on a payload that will travel into space this fall. The payload is a digital twin that will use AI software to measure the activity and predict the future state of the battery. (CREDIT: Mario Rodriguez)

A Smart Brain for the Satellite

The most exciting feature of this mission is the custom payload—a special package inside the satellite built by researchers. This package holds a digital twin, which is a computer model that acts like a mirror of the satellite’s power system. But unlike earlier versions of digital twins that stay on Earth and get updates sent from space, this one lives and works inside the satellite itself.

That means the satellite doesn’t need to “phone home” to understand how it’s doing. Instead, it uses built-in sensors and software to constantly check the health of its own batteries, monitor power levels, and decide what might happen next.

“The spacecraft itself can let us know how it’s doing, which is all done by humans now,” explained Adam Zufall, a graduate researcher helping to lead the project.

By using artificial intelligence, the satellite’s brain doesn’t just collect data. It also learns from it. Over time, the system should get better at guessing how its batteries and systems will behave next. That helps the satellite adjust its operations on its own, even before problems arise.

“It should get smarter as it goes,” said Professor Stephen Robinson, who directs the lab that built the payload. “And be able to predict how it’s going to perform in the near future. Current satellites do not have this capability.”

Working Together Across Disciplines

Building this kind of technology takes teamwork. The project brings together experts in robotics, space systems, computer science, and battery research. In addition to Robinson and Zufall, the team includes another mechanical engineering professor who focuses on battery management. His lab studies how batteries behave under different conditions, including in space.

A satellite made by Proteus Space with payload help from UC Davis researchers, is planned for an October launch that the aerospace company says would make it the fastest-ever launch-qualified satellite. (CREDIT: Proteus Space)

Graduate students in engineering and computer science also play major roles. One student helps design the spacecraft’s software, while others work on how the AI makes predictions and responds to changes in power levels.

Together, they’ve built a model that can monitor voltage and other readings to understand how much energy the satellite can store and use.

The satellite will carry several other payloads, both commercial and scientific. But the real highlight is this AI-powered system that watches itself and adjusts on the fly.

What Happens After Launch

Once launched from Vandenberg Space Force Base, the satellite will move into low Earth orbit. It’s designed to stay active for up to 12 months, gathering data and testing its smart brain in the harsh conditions of space. This type of orbit sits a few hundred miles above the Earth’s surface—far enough to test the systems, but close enough for short communication times.

The Proteus team just broke the record for the fastest launch qualified custom bus satellite ever. (CREDIT: Proteus Space)

After its mission ends, the satellite will continue to orbit for another two years. By the end of its life, gravity and drag will pull it back toward Earth, where it will burn up safely in the atmosphere. This kind of planned decay helps keep space clean and reduces the risk of debris collisions.

The whole mission shows how fast and flexible future space projects might become. Instead of waiting years to build and test systems, researchers could design, launch, and operate smart satellites in a matter of months. That could open the door to more frequent missions, more advanced designs, and smarter satellites across the board.

Changing the Future of Spacecraft

Satellites that can take care of themselves offer big advantages. Right now, spacecraft rely on ground teams to tell them what to do, run checks, and respond to problems. This creates delays, increases costs, and adds risk.

Another angle of the satellite. (CREDIT: Proteus Space)

By placing real-time digital twins on board, future satellites could adjust to problems on their own. They could shut down failing parts, save power when needed, or warn engineers of upcoming issues days in advance.

This would reduce the workload for ground teams and improve the life and safety of space missions.

The team behind this project believes their work is just the beginning. With more advanced AI tools and faster build times, space technology could move at a much quicker pace. More importantly, it could become smarter, more reliable, and more responsive to change. This satellite might be small, but it could help start a big shift in how space systems are built and run.

Source link