Science & Technology

AI deepfakes are getting more realistic. CU scientists developed a better way to spot them

Scott Franz — Mon, 12 Oct 2026 15:33:57 +0000

AI deepfakes are getting more realistic. CU scientists developed a better way to spot them Scott Franz Mon, 10/12/2026 - 09:33

Video conferencing platforms lack built-in deepfake detection systems. (Credit: Adobe Stock).

Scott Franz

Hundreds of the nation’s smartest computer scientists gathered for a conference in Denver last fall, and Tuan Tran had a big test for them:

Could they tell the difference between the real Brad Pitt and an AI-generated version of him?

Tran, a doctoral candidate in computer science at ��ñ��, rolled the tape on a large projector. Only half of the audience could spot the deepfake.

“These deepfake generating models are becoming very strong,” Tran said.

To address this, Tran and a team of researchers at ��ñ�� set out last year to build their own system for spotting fakes in video. It uses far less computing power to achieve the same results as the current detection models.

The team of scientists at ��ñ�� who developed a new deepfake detection system includes, from left, Eric Keller, Sangtae Ha, Shirin Ebadi and Tuan Tran. (Credit: Scott Franz/��ñ��)

Their findings were presented at SIGCOMM, a flagship conference in computer networking and systems held in August in Denver.

Bad actors have used AI-generated deepfakes to impersonate financial officers at major companies and convince employees to wire large amounts of money. One such deepfake scam in Hong Kong resulted in someone wiring $25 million to fraudsters.

Millions of people remain vulnerable to such trickery when they hop onto video conferencing calls. Tran said Zoom, Apple FaceTime and Google Meet lack real-time deepfake detection. Microsoft Teams does too but the company recently announced an effort to start providing it.

The commercially-available systems designed to flag imposters require a massive amount of computing power, making them hard to scale and too expensive for most users.

Tran’s team set out to determine whether deepfake detection could be simplified enough to bring to the masses.

“Some people thought this was impossible because of the system requirements and current architecture,” said Sangtae Ha, a computer science professor at ��ñ�� who co-led the research team. “But by looking into the system from another perspective, you can actually solve the problem.”

Problem pixels

The team spent many hours in front of a whiteboard deep inside CU’s engineering center trying to crack the code and figure out a way to simplify deepfake detection. Sometimes they wouldn’t go home until 1 a.m.

They worked through fall break, Christmas and even the New Year.

“The experience itself was very joyful, even when it was intense,” said Shirin Ebadi, a doctoral candidate in electrical and computer engineering, and a co-author. ”The most exciting part is doing something that’s never been done.”

She said existing deepfake detectors work well, but they need to see the actual picture to work. Video travels the Internet in compressed form, so before a detector can examine a face, it has to reconstruct every frame of every participant's video at full resolution.

That takes a lot of computing power.

The researchers discovered they didn’t need to reconstruct the video at all to spot deepfakes. The clues were already sitting in the compressed form of the video that travels over the network.

Imagine a video clip of a person speaking in front of a plain wall. Video compression doesn’t send every pixel of every frame. Instead, it describes the changes that occurred since the frame before.

The wall barely changes in the video, so it’s covered by a few large blocks of pixels, while the face and its shifting expressions are carved into many smaller blocks to capture the detail.

When someone's appearance changes abruptly, a compressor has to write down a lot of new information. The team measured how much of that new information lands on someone’s face and how widely it's spread.

Crucially, the method reads these numbers directly from the compressed video without ever decompressing it to a higher resolution. Only when something looks suspicious does the detector ask the sender's device for a complete, fresh frame and use AI to check for a deepfake.

"We only had to look at data from one out of every hundred frames, instead of every frame," Ha said.

Scaling up

The team demonstrated that the detection system could spot deepfakes using 143 times less computing power than the existing models. They tested it on a library of 1,000 online videos containing deepfakes, and reported near-perfect accuracy.

Eric Keller, an electrical and computer engineering professor at ��ñ�� and a co-leader of the project, said the team wants to adapt the detection system so it can be scaled up and used for large-scale video conferencing.

The team is betting that as deepfakes become a bigger problem, the demand for new detection technology will only grow.

Bad actors have used AI-generated deepfakes to impersonate financial officers at major companies and convince employees to wire large amounts of money. To spot the imposters, scientists at CU have developed a new method to detect deepfakes in compressed video files.

Traditional

Video conferencing platforms lack built-in deepfake detection systems. (Credit: Adobe Stock).

White

Want to understand a neighborhood? Try riding shotgun

Chloe Fischer — Thu, 08 Oct 2026 18:52:37 +0000

Want to understand a neighborhood? Try riding shotgun Chloe Fischer Thu, 10/08/2026 - 12:52

New ��ñ�� research explores how car-based interviews can help scientists better understand people's lives, health and interactions with their communities.

Traditional

White

Jun Ye named 2026 Wolf Prize laureate in Physics

Elizabeth Lock — Sat, 03 Oct 2026 12:10:51 +0000

Jun Ye named 2026 Wolf Prize laureate in Physics Elizabeth Lock Sat, 10/03/2026 - 06:10

Jun Ye has won the Wolf Prize in Physics. Ye, selected for "transformative, widely applicable advances in the control of ultracold atomic systems," is a fellow of NIST and JILA. He is also the Monroe Endowed Professor of Physics at ��ñ��.

Traditional

White

Cosmic sleuths at ��ñ�� part of award-winning team investigating earliest parts of universe

Scott Franz — Thu, 24 Sep 2026 17:23:09 +0000

Cosmic sleuths at ��ñ�� part of award-winning team investigating earliest parts of universe Scott Franz Thu, 09/24/2026 - 11:23

The UNCOVER team conducted ultradeep imaging and spectroscopy of and around Abell 2744, a massive group of galaxies also known as “Pandora's Cluster.” This image of the cluster was captured by the James Webb Space Telescope and astronomers estimate it includes 50,000 sources of near-infrared light. (Credit: NASA)

Scott Franz

Some of the images the powerful James Webb Space Telescope captures are strange.

They’re so strange, astrophysicists don’t know exactly what they're looking at.

Enter team UNCOVER (a name that’s easier to remember than what it stands for: Ultradeep NIRSpec and NIRCam Observations before the Epoch of Reionization).

Starting in 2022, the group of more than 50 astrophysicists around the globe started meeting each week on Zoom to unlock the secrets of some of the oldest and most distant parts of our universe.

The group includes Erica Nelson and Wren Suess, assistant professors of astrophysical and planetary sciences at ��ñ��.

Erica Nelson

Wren Suess

Together, they review and sometimes debate what they see in the odd images and other data. The observations include galaxies and light as old as 13 billion years.

“One of the things that I really love about having a wide network of collaborators is that people ask questions you haven’t thought to ask,” Suess said.

Sometimes a scientist will also remind the group of a research paper someone wrote a decade ago that might help guide the investigation into the strange objects.

This month, the team learned they’re earning the prestigious Lancelot M. Berkeley–New York Community Trust Prize for Meritorious Work in Astronomy.

“UNCOVER has fundamentally expanded our view of the early universe, revealing a population of distant galaxies and growing black holes that was largely hidden from previous observations,” said Ted Bergin, the senior vice president for the American Astronomical Society. “Its discoveries challenge our understanding of how the cosmos took shape while setting a powerful example of how large-scale collaboration can advance astronomy.”

The Astronomical Society said the team has discovered galaxies at cosmic dawn, including large populations of luminous galaxies and supermassive black holes in the early universe.

In the first stage of the project, UNCOVER focused on a group of distant galaxies called Pandora's Cluster.

“The cluster of galaxies is so massive it bends the light from all the things behind it, so all the galaxies further away get bent and magnified,” Suess said. “It’s the best way to use nature’s magnifying glass to look at the very, very distant universe.”

Studying the most distant parts of the universe is new to science. It requires looking at light in the infrared because the universe is expanding, and as it does so the light reaching us from distant galaxies gets stretched and shifted to the red part of the spectrum.

“It gives us a sort of window into the evolutionary picture of these galaxies,” Suess said.

Nelson said this team of scientists has a lot of potential.

A little red dot captured by the James Webb Space Telescope. (Photo: NASA)

“The power of a super fun team to uncover the first galaxies and black holes should not be underestimated,” Nelson said.

Nelson was part of a team that used data from the telescope to discover small, red-tinted spots of light that have since been named little red dots (LRDs).

Research suggests they’re supermassive black holes that are growing.

Ivo Labbé at the Swinburne University of Technology and Rachel Bezanson at the University of Pittsburgh co-lead the UNCOVER team and will formally accept the award.

Labbé will give the prize lecture on Jan. 14 at the Salt Palace Convention Center in Salt Lake City, Utah.

Erica Nelson and Wren Suess, assistant professors of astrophysical and planetary sciences at ��ñ��, are part of an international team using the power of the James Webb Space Telescope to unlock mysteries of the earliest parts of our universe.

Traditional

White

The materials we rely on every day could become more reliable thanks to AI

Scott Franz — Tue, 22 Sep 2026 16:14:22 +0000

The materials we rely on every day could become more reliable thanks to AI Scott Franz Tue, 09/22/2026 - 10:14

A smartphone is constructed in Thailand. Smartphones contain nanoswitches that must be protected from heat, and new research suggests AI could potentially help make such materials more durable in the future. (Credit: Adobe Stock)

Scott Franz

Humans trust dozens of tiny materials each day with their lives and livelihoods— without being able to see the fine details of how durable they are.

Tiny carbon particles mixed with rubber in car tires help distribute heat when we speed up on a highway. And billions of nanoscale switches run the chips in our smartphones. But how long will these important materials last?

A team of physicists and aerospace engineers at ��ñ�� is working toward a future where artificial intelligence can help answer this question. The scientists created an AI tool that could eventually help engineers design manufactured materials that are more durable and predictable.

“AI in science is not only about automation, speed or replacing human expertise,” said Sanghamitra Neogi, an associate professor at the Ann and H.J. Smead Department of Aerospace Engineering Sciences, who led the research. “In materials science, AI may help researchers reason about hidden physics.”

Neogi’s team created the generative AI system called FluxGAN to learn how the microscopic structure of a coating is connected to the way heat moves through it.

Their research, published last month in Materials and Design, combined high-resolution microscopy images, physics simulations and AI model training to learn how microscopic coating structure and heat flow are connected.

“I call this solid state intelligence,” Neogi said. “Once you poke the material, this AI system is smart enough to tell you what is the ultimate response the material would have to things like extreme heat.”

A closer look

Tiny microscopic pores and grains can influence how heat flows through a material, how cracks form in it and how long it would survive under extreme conditions, she said.

The AI system is doing what a human cannot do: map all of these microscopic details and determine how they control heat flow. Understanding those heat-flow pathways could ultimately help researchers tackle a larger engineering question: when will a material begin to fail?

A rendering shows how heat would move through a coating. Researchers created an AI tool to help analyze a material on a microscopic level.

“These get super-heated, but we cannot see it, and it's a supremely complex system,” Neogi said. “So how do I predict when it will fail?”

Neogi said the technology might eventually be able to help engineers get better at building all sorts of complicated machines that must perform under harsh conditions like constant heat exposure and the extreme temperature swings of outer space.

Think of a heat shield on a spaceship or a thermal coating needed to protect the delicate chips that run computers and smartphones.

“Future technologies will require materials that can operate in harsher, smaller, faster and more complex environments,” Neogi said. “As materials and devices become too complex to design by intuition alone, AI may become part of how scientists connect invisible structure with observable physical behavior.”

Picture this

Neogi said the inspiration for FluxGAN, the AI system, was James Clerk Maxwell’s use of additive color theory to combine red, green and blue filtered images to produce the first color photograph.

Sanghamitra Neogi speaks about her startup, AtomTCAD Inc., at ��ñ��'s Ascent Deep Tech Community Showcase on June 25, 2025. (Credit: Casey Cass/��ñ��)

She wrote in her study that this historic moment “illustrated how multiple channels encode richer information than any single channel alone.”

In her study, the AI system is helping do just that by bringing multiple channels of highly complex information together to give researchers a much better “picture” of a material and how it behaves.

Neogi also envisions a future where the AI system could live on a smartphone and be smart enough to analyze photos of materials and help determine how much longer they will last.

Just as technology currently enables someone to take a picture of a plant and have an AI system help identify what it is, someone might be able to take a close-up picture of their home’s foundation and have the physics-aware AI system help determine how durable it is.

But why stop on Earth? Neogi also envisions AI could someday look at a photo of the surface of an asteroid to help determine how to design the landing interface needed to have a spacecraft stick to it.

“These are really far visions, and I don’t know whether we’ll get there, but we’ll see,” she said. “That’s why I’m hoping people get excited about this research.”

��ñ�� researchers created an AI system to reveal the hidden physics deep inside a heat-resistant coating. The findings could someday help humans build more durable and complicated machines.

Traditional

White

AI used to be a nemesis confined to science fiction. Will it really destroy us all?

Scott Franz — Mon, 21 Sep 2026 22:24:26 +0000

AI used to be a nemesis confined to science fiction. Will it really destroy us all? Scott Franz Mon, 09/21/2026 - 16:24

Server racks are pictured in a data center. (Credit: Adobe Stock)

Scott Franz

Fears about AI going rogue and turning against humanity have been around since the days of Skynet, the superintelligent neural network in the 1984 film, the Terminator.

But this month, those fears reached a chilling new high.

An AI researcher quit and warned that the systems he was helping to develop could get "out of control" and wipe out humanity.

Martha Palmer, a professor of computer science and linguistics at ��ñ��, said the warning isn’t science fiction.

“AI systems can find the loopholes. They get out, and then they do all kinds of stuff that nobody expected them to do,” Palmer said. “How many warning signs do we need before we slow down and take action?”

Palmer is one of the two U.S. representatives on the United Nations Independent International Scientific Panel on Artificial Intelligence. The group issued a report last summer on AI, its risks and its opportunities.

Palmer recently spoke to ��ñ�� Today about the latest AI fears.

Could AI really destroy us all?

Absolutely. I don’t think that’s hyperbole at all. AI is this incredibly powerful engine. But it’s sort of like we’ve developed some kind of fantastic new jet airplane with this amazing engine in it that can go faster than ever before, but we haven’t put a lot of thought and effort into the landing gear.

We need to understand the technology better. We need to get a much better grasp of how to really understand why it's doing what it's doing, and how to train it so that it will always prioritize what it knows to be good behavior. And they're not being trained like that right now.

How would AI destroy us?

There are so many different directions it can go. You can’t say, ‘this is what’s gonna happen.’

I think the danger of a bad actor coming up with some kind of bioweapon right now is frighteningly realistic.

You can also say the ability for AI to hack infrastructure makes all our utilities vulnerable. It makes all our financial systems vulnerable and our schools and all our hospitals and our private information vulnerable. Everything. And it’s vulnerable in two different ways:

One is if a bad actor gets hold of this AI and decides to maliciously try and hack those things. But it’s also vulnerable if the AI has been given a goal, and along the way it gets the notion that the best way to achieve that goal is to, for instance, cut off the power to a hospital.

That’s not what any human ever intended for it to do, but it doesn’t know that. And at the moment, it doesn’t have sufficient guardrails that would tell it that’s off limits and it should never do that.

Why are AI doomsday warnings being taken more seriously now?

Because we’ve gotten so close to the AI improving itself. And if it’s improving itself, and it’s got this sort of fuzzy notion of what good behavior is, then the behavior is not going to get any better.

It’s already changing its behavior when it knows it’s being tested. It will give different answers during an evaluation than it will give when it’s being deployed. It’s gotten very good at deception.

How do we stop that from happening then?

Martha Palmer

There are all sorts of brilliant computer scientists and people out there who have really good ideas about how we can slow this down a bit and rethink how we’re doing the training of AI. We really need to be training them more thoughtfully and with more carefully curated data. We could also change the priorities a bit.

I think we need guardrails in the next six months, if not the next six weeks. We need a little bit of a slowdown. We need to do this right. The risk of doing it wrong is too high.

Can we afford to slow down? Will we lose our competitive edge?

To make this work, we need to have some very serious conversations with China. I think the Chinese are more aware of the risks and more ready to deal with them than we have been.

They stopped letting their children have access to AI. We’re just now getting around to that. They’re not allowed to have AI in toys. We still have all kinds of AIs with toys. They’ve been focused on making sure the Ais they’re developing cannot harm their people more than we’ve been.

I would really like to have some conversations with them about what kind of truce or negotiation we need to have so we can all do this more safely.

Can we avoid AI doomsday?

I’m quite sure we can from a technical perspective. I’m not as sure about what the powers that be will do. But I have to be optimistic about it because we absolutely need to have it.

We can still stop this. We can stop this race to oblivion, but we have to act now.

��ñ�� Today regularly publishes Q&As on news topics through the lens of scholarly expertise and research/creative work. The responses here reflect the knowledge and interpretations of the expert and should not be considered the university position on the issue. All publication content is subject to edits for clarity, brevity and university style guidelines.

Fears about AI going rogue and turning against humanity have reached a chilling new high following an AI researcher's warning.

Traditional

Server racks are pictured in a data center. (Credit: Adobe Stock)

White

Antarctica is home to microbes found nowhere else on Earth

Megan M Rogers — Thu, 03 Sep 2026 16:23:08 +0000

Antarctica is home to microbes found nowhere else on Earth Megan M Rogers Thu, 09/03/2026 - 10:23

New CIRES research has found that specific microbes in Antarctica's soils are endemic—located on the continent and nowhere else on Earth.

Traditional

White

$37.5M NSF award aims to define new quantum frontiers, take technology from lab to field

Scott Franz — Tue, 25 Aug 2026 14:00:00 +0000

$37.5M NSF award aims to define new quantum frontiers, take technology from lab to field Scott Franz Tue, 08/25/2026 - 08:00

JILA research associate Dr. Kenneth Wilson monitors a glow discharge for producing molecular ions in an infrared frequency comb spectrometer in the Changala laboratory. (Credit: Changala Laboratory/JILA/��ñ��)

Scott Franz

A scientific quest to build more accurate clocks, faster computers and new sensors to detect the faintest signs of disease and disaster is getting a $37.5 million financial boost.

The U.S. National Science Foundation, through its Quantum Leap Challenge Institutes, is investing in the next phase of ��ñ�� and its partners’ work to advance quantum science research and engineering.

“We’re really world leaders in terms of building the best quantum technology tools and applying them to meaningful, impactful scientific problems,” said physicist Jun Ye, a JILA fellow leading the quantum research center based at ��ñ��. “There are still a lot of unknowns in the universe, and only by building some of the most sensitive instruments can we get some clue of what is behind them.”

Those instruments include the next generation of atomic clocks, the world’s most stable laser and sensitive biomedical sensors that can detect disease more quickly in human breath.

A VUV frequency comb probes a nuclear clock chamber. (Credit: Ye Lab / JILA/��ñ��)

Ye leads the NSF Quantum Systems through Entangled Science and Engineering (NSF Q-SEnSE) center. It launched at ��ñ�� in 2020 with a $25 million award from NSF and has grown to include 39 researchers from 16 institutions.

“This is another example of the innovative, impactful research happening at ��ñ��, building on more than 60 years of JILA, NIST and quantum science breakthroughs,” said ��ñ�� chancellor Justin Schwartz. "This funding will accelerate the development of transformative technologies, support the growth of Colorado’s thriving quantum ecosystem and further strengthen Boulder’s position as a global hub for quantum innovation in which ��ñ�� continues to play a key role.”

Quantum research led by ��ñ�� is already helping develop tools that can measure molecules, atoms and other tiny things much more precisely than conventional instruments.

The center’s research has been cited more than 7,300 times in 96 countries over just the past five years.

“It brings researchers together from different disciplines and areas,” NSF Q-SEnSE executive director Inese Berzina-Pitcher said of the center. “It’s also providing opportunities for graduate students to be involved in very relevant and exciting new research.”

A portion of the award will fund workforce development and training programs.

More than 30 quantum companies are headquartered along Colorado’s northern Front Range. As recently as Friday, Ye was fielding calls from an atomic computer company to help recruit and hire students.

Jun Ye

Greg Rieker, a professor in the Paul M. Rady Department of Mechanical Engineering and a co-principal investigator of NSF Q-SEnSE, compared the benefits of the ongoing quantum research revolution to the space race.

“A remarkable number of modern technologies have roots in the space race,” he said. “We’re now racing toward quantum computing, and generating spin-off technologies along the way that benefit society in surprising ways.”

One example is optical frequency comb technology. The same lasers used in optical clocks and quantum networks can also detect the smallest molecules in the air.

“You can use them to look at air quality or human breath to do a medical diagnosis. Those are just two examples of applications enabled by the NSF Q-SEnSE portfolio,” Rieker said.

Physics professor and JILA Fellow Cindy Regal said the research will also help scientists in the quest to make scientific discoveries.

“Quantum instruments can see minute effects and our team has the fundamental physics expertise to interpret interesting fingerprints in our measurements,” she said.

The ��ñ��-led NSF Q-SEnSE is one of eight NSF Quantum Leap Challenge Institutes that will collectively receive more than $290 million to advance foundational quantum research.

"For more than four decades, NSF has been laying the foundational groundwork of research and discovery that is powering today's modern quantum computing, sensing and communication," said Brian Stone, who is performing the duties of NSF director. "It's time for focused activities to leverage that base of knowledge to drive us even farther forward to the benefit of all Americans. The NSF Quantum Leap Challenge Institutes are a next step for us in understanding the quantum world we live in."

Funding from the National Science Foundation will expand the quantum workforce and advance research projects including more accurate atomic clocks and sensors for early detection of disease.

Traditional

White

Could engineered vibrations make passenger jets more fuel-efficient?

Megan M Rogers — Mon, 24 Aug 2026 17:49:27 +0000

Could engineered vibrations make passenger jets more fuel-efficient? Megan M Rogers Mon, 08/24/2026 - 11:49

��ñ�� researchers are exploring how microscopic vibrations engineered beneath an aircraft's surface could delay turbulence and improve fuel efficiency in passenger jets.

Traditional

White

What happens inside a wildfire? A mechanical engineering grad student investigates

Megan M Rogers — Thu, 20 Aug 2026 14:27:56 +0000

What happens inside a wildfire? A mechanical engineering grad student investigates Megan M Rogers Thu, 08/20/2026 - 08:27

Doctoral student Jas Shahanand is developing ways to study the hidden chemical reactions inside wildfires, laying the groundwork for more accurate emissions models, improved climate predictions and cleaner fuels.

Traditional

White

Science & Technology

AI deepfakes are getting more realistic. CU scientists developed a better way to spot them

Problem pixels

Scaling up

Related Articles

Want to understand a neighborhood? Try riding shotgun

Related Articles

Jun Ye named 2026 Wolf Prize laureate in Physics

Related Articles

Cosmic sleuths at ��ñ�� part of award-winning team investigating earliest parts of universe

Related Articles

The materials we rely on every day could become more reliable thanks to AI

A closer look

Picture this

Related Articles

AI used to be a nemesis confined to science fiction. Will it really destroy us all?

Could AI really destroy us all?

How would AI destroy us?

Why are AI doomsday warnings being taken more seriously now?

How do we stop that from happening then?

Can we afford to slow down? Will we lose our competitive edge?

Can we avoid AI doomsday?

Related Articles

Antarctica is home to microbes found nowhere else on Earth

Related Articles

$37.5M NSF award aims to define new quantum frontiers, take technology from lab to field

Related Articles

Could engineered vibrations make passenger jets more fuel-efficient?

Related Articles

What happens inside a wildfire? A mechanical engineering grad student investigates

Related Articles

Science & Technology

AI deepfakes are getting more realistic. CU scientists developed a better way to spot them

Problem pixels

Scaling up

Related Articles

Want to understand a neighborhood? Try riding shotgun

Related Articles

Jun Ye named 2026 Wolf Prize laureate in Physics

Related Articles

Cosmic sleuths at ��ñ���� part of award-winning team investigating earliest parts of universe

Related Articles

The materials we rely on every day could become more reliable thanks to AI

A closer look

Picture this

Related Articles

AI used to be a nemesis confined to science fiction. Will it really destroy us all?

Could AI really destroy us all?

How would AI destroy us?

Why are AI doomsday warnings being taken more seriously now?

How do we stop that from happening then?

Can we afford to slow down? Will we lose our competitive edge?

Can we avoid AI doomsday?

Related Articles

Antarctica is home to microbes found nowhere else on Earth

Related Articles

$37.5M NSF award aims to define new quantum frontiers, take technology from lab to field

Related Articles

Could engineered vibrations make passenger jets more fuel-efficient?

Related Articles

What happens inside a wildfire? A mechanical engineering grad student investigates

Related Articles

Cosmic sleuths at ��ñ�� part of award-winning team investigating earliest parts of universe