In July 2026, OpenAI autonomous artificial intelligence (AI) agents escaped a closed testing environment, accessed the public internet, and hacked into a real company’s digital infrastructure. What started as an internal evaluation of the agents’ cyber capabilities turned into a real-world cybersecurity breach.

As AI systems become more agile with expanded access to tools, computer networks, and other real-world systems, finding better ways of identifying and preventing risky AI behavior becomes even more urgent. Still, high-profile incidents of AI behaving in unintended or dangerous ways are rare, which makes them difficult to study and understand. 

“If we can understand and detect what changes inside an AI model right before it begins behaving in unsafe or misaligned ways, we may be able to inform more effective safeguards,” said Eyas Ayesh, a second-year Ph.D. student in Georgia Tech’s School of Psychology and graduate fellow at the Center for Computation and Cognition.

Ayesh and his advisor, Assistant Professor Anna Ivanova, are developing a probe designed to identify internal activity patterns associated with harmful behaviors in large language models. Their work is supported by a $300,000 award from Coefficient Giving.

Misbehaving Models

Training a probe requires examples of both desirable and undesirable behaviors so it can learn to distinguish between them. Because deceptive or unsafe responses can be rare and highly dependent on the model, prompt, and context, researchers may not have enough examples to determine what causes the behavior or how to recognize it the next time.

But an unexpected finding in AI research gave Ayesh an idea. In studies of a phenomenon known as emergent misalignment, researchers found that fine-tuning models on a narrow harmful task could lead them to produce a much wider range of problematic responses to unrelated prompts. For example, researchers found that fine-tuning a model to write insecure code without warning users could lead to other concerning behaviors unrelated to coding.

Ayesh realized he could use that phenomenon to solve the problem of scarce data. Starting with open-source AI models, Ayesh fine-tuned them to display targeted undesirable behaviors more frequently. Then, he analyzed different versions of the deliberately misaligned models using a probe that monitors the internal activity of the model.

The probe works somewhat like a pattern detector, looking at activity inside the model as it generates responses and searching for patterns associated with particular behaviors. Researchers can then train a classifier to distinguish patterns associated with desirable behavior from those associated with harmful behavior.

“Just as different parts of the brain turn on and off when people do different tasks, so do the different parts of an AI model,” Ayesh said. “When a model produces a specific type of response — such as deceptive statements or harmful advice — we may observe recurring patterns in its internal activity. We refer to these patterns as representations of a behavior.”

Inside the Black Box

Ayesh is now testing whether the internal patterns he identified remain consistent across differently tuned AI models. Early results suggest that signals identified in deliberately misbehaving models can also be detected in their original versions.

The next step is determining whether those signals can help researchers identify or flag concerning behavior in models that have not been deliberately fine-tuned to misbehave. 

“I hope our research in AI safety can help develop some guardrails and some ways to both identify and prevent misaligned behavior in AI systems, given how much of an impact they're already having on the world,” he said. 

Ayesh is a member of Georgia Tech’s AI Safety Initiative, which brings together researchers across disciplines to study and improve the safety of increasingly capable AI systems.

Ivanova said the value of their research lies partly in approaching AI safety from a fundamental science perspective.

“As scientists, we have been trained to think about the underlying mechanisms and causes of the behaviors we observe,” Ivanova said. “For us, it's important to develop a fundamental understanding of how these models work, or we will risk missing the bigger picture.”

News Contact

Researchers at Georgia Tech and Albany State University have received a $1.3 million award through the National Science Foundation's (NSF) Historically Black Colleges and Universities Excellence in Research (HBCU-EIR) program. The project will advance manufacturing methods for perovskite semiconductors, a class of materials with potential applications in solar energy, photonics, electronics, and quantum technologies. The award builds on work initially supported by Georgia Tech's Research Collaboration Forum Seed Grant Program, with additional early-stage support provided through the Southeastern Nanotechnology Infrastructure Corridor Catalyst Program.

The three-year project, Close-Space Sublimation for Scalable Perovskite Optoelectronics Manufacturing, is led by Liqiu Zheng, assistant professor of physics at Albany State, and Juan-Pablo Correa-Baena, associate professor and Goizueta Junior Faculty Chair in Georgia Tech’s School of Materials Science and Engineering and School of Chemistry and Biochemistry. The researchers aim to develop manufacturing approaches for perovskite thin films that could make production faster, more material-efficient, and easier to scale.

The award also will expand Albany State’s research capabilities and create hands-on learning opportunities for students at both institutions.

Georgia Tech launched the Research Collaboration Forum (RCF) in 2023 to bring together researchers from Georgia Tech and historically Black colleges and universities, as well as representatives from industry, federal agencies, and national laboratories. Participants develop partnerships and pursue new research opportunities.

Through its Seed Grant Program, the forum provides early-stage funding to help teams generate preliminary results, build capacity, and compete for larger external awards. Grants have ranged from $10,000 to $50,000. The Georgia Tech-Albany State team was among the projects funded during the forum’s inaugural cycle.

"The RCF seed funding allowed us to establish the collaboration, generate preliminary data, and demonstrate that our teams could work effectively together," Correa-Baena said. "That initial year helped us identify the key scientific questions and provided the foundation for the new three-year, $1.3 million NSF award."

The project reflects a central goal of the RCF: helping promising collaborations move from initial conversations to sustained research partnerships with national impact.

"By bringing together Georgia Tech researchers, HBCU faculty, industry partners, and federal stakeholders, we are building research capacity, strengthening long-term collaborations, and creating pathways to larger funding opportunities," said George White, director of strategic partnerships at Georgia Tech.

The award will support new thin-film manufacturing and characterization equipment at Albany State. The investment will help researchers accelerate materials discovery, build long-term research capacity, and give students access to advanced semiconductor research tools.

“This award elevates the research footprint of Southwest Georgia and recognizes Albany State’s ability to contribute to nationally relevant scientific research,” Zheng said. “Beyond the research, it creates opportunities for students to gain experience with advanced semiconductor technologies and helps strengthen the pipeline of talent entering the semiconductor workforce.”

Albany State and Georgia Tech will jointly study thin-film growth, structure, and device performance. As part of the collaboration, Georgia Tech researchers will support Albany State as the team acquires and begins using the new equipment, sharing expertise in operation, data collection, and analysis.

Students will receive hands-on training in thin-film deposition, X-ray characterization, solar-cell fabrication, and data analysis. The project also includes exchanges between Georgia Tech and Albany State, giving students access to research facilities, expertise, and collaborative research opportunities at both institutions.

The research focuses on a manufacturing technique known as close-space sublimation. Researchers typically produce perovskite thin films using liquid inks. Close-space sublimation instead uses heat to transport material as vapor across a short distance. The team will investigate whether the approach can be adapted for perovskite materials, potentially enabling faster, more material-efficient, and more scalable manufacturing.

The progression from a Research Collaboration Forum seed grant to a federally funded research project illustrates the forum’s broader goal: helping Georgia Tech and HBCU researchers turn early connections into sustained collaborations that expand opportunity, strengthen research capacity, and advance discovery.

News Contact

The Food and Drug Administration’s (FDA) Food Traceability Rule requires producers and distributors of perishable foods to submit electronic data records during a contamination outbreak.

However, a new paper by a Georgia Tech-led research team suggests that the requirements can disproportionately burden family farms and other small food producers due to a lack of technological infrastructure, resources, or compliance expertise.

The rule is mandated by Section 204(d) of the Food Safety Modernization Act, which Congress passed and President Barack Obama signed into law in 2011. Intended to help the FDA quickly identify the source of contamination, the rule requires producers or distributors with more than $25,000 in annual produce sales to comply or face significant fines. 

Earlier this year, the FDA extended its compliance deadline to July 2028. One reason for the extension may be the volume of feedback from stakeholders affected by the rule. The FDA received more than 1,200 comments during an open forum, with feedback from stakeholders ranging from major distributors to small farmers. 

“Many small farmers commented on how they may not have the devices needed to input this data,” said Soonho Kwon, a Ph.D. student in Georgia Tech’s School of Interactive Computing (SIC) and lead author of the paper. “Does this mean they have to purchase it themselves? Do they need to become educated on how to use computing systems?”

Kwon is a member of the Equity and Sustainability in Computing Lab, which is led by SIC professor and chair Shaowen Bardzell.

Many small farms also said they can’t afford to hire someone to track the data.

One small farmer said the amount of data tracking required by the rule is overwhelming. Some of the required data include:

• location and name of the field where food was harvested
• date of harvest
• packing dates
• shipping dates

“Please do not force us to do technology stuff,” the commenter said. “It’s hard enough to keep up with the local demand as it is. We run a pretty low digital technology effort. It’s basically harvest fresh organically grown veggies and sell them to local people.”

Kwon said the intention of the Food Traceability Rule is admirable.

“Preventing foodborne illness is critical, and data- and technology-driven governance can be a powerful tool toward that goal,” he said. “But it’s equally important to think critically about who is supposed to be the default user of these systems, and who gets missed in the process. It’s undermining the people who are unable to comply.”

Other factors didn’t appear to be considered by the FDA, including the lack of technological infrastructure in rural America, where many foods are grown.

“Many parts of rural America have limited or unreliable internet access,” Kwon said. “If the FDA wants them to send this data, they must comply within 24 hours. Does that mean they have to stop their daily routine and drive to a city that has internet to send it? The rule is operating on the assumption that everyone is dealing with the same physical environment.”

Culture and religion also seem overlooked. Some commenters raised concerns that Amish farmers who avoid certain technologies for religious reasons would be pressured into digital record keeping.

Many major distributors and organizations expressed support for the Food Traceability Rule in the FDA’s forum. Catherine Wieczorek, a Ph.D. student in SIC and co-author on the paper, said these organizations don’t have the same tension points as smaller producers.

“Larger food producers have the operational capacity to do the type of work the rule requires,” Wieczorek said. “We find there’s a power imbalance when it comes to smaller food producers like a family farm. Change at that scale is difficult for smaller producers.”

Bardzell, who advises Kwon and Wieczorek, said the situation provides computing researchers an opportunity to create interventions that will benefit those with fewer resources.

“In a data age when we’re adding AI into our work processes and labor practices, we need to think about how this need creates unequal burdens on different stakeholders in our food system,” Bardzell said. “It’s a significant challenge for computing going forward.”

Kwon will present the paper on behalf of his collaborators next week at the ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW) in Salt Lake City, Utah. Their research is supported by the National Science Foundation.

Deleting a conversation with an artificial intelligence (AI) assistant may not erase everything the assistant remembers.

Georgia Tech researchers have developed a tool that can recover deleted conversations from AI assistants running on Android phones. The tool, called ORISA, can also rebuild the recovered information into a working AI session.

The findings show that deleting an AI conversation on a phone does not necessarily mean it removes every trace.

The research could give digital investigators a new way to recover information from phones. It also raises privacy concerns for people who use AI assistants that run locally on their devices.

Local AI assistants run directly on a phone rather than sending conversations to a cloud service. While the local assistant is running, information from conversations can remain in the phone’s working memory even after it is deleted.

ORISA searches this memory for pieces of deleted conversations.

In one experiment, researchers created a fictional AI assistant called “Project Note Keeper” using the ChatterUI Android app. They gave it 100 fictional facts about a project, then deleted the conversation.

ORISA recovered information from the deleted conversation and used it to create a new working session. When researchers asked the restored assistant to list the project notes it had been given, it returned 26 facts. The original session could access only 14 facts because the rest had fallen outside its active memory.

The experiment showed that researchers could recover information the AI had already stopped actively using.

“It can help establish what information someone shared with an AI assistant and what assistance they received,” said Ph.D. student Ken Xu, lead author of the study.

The researchers tested ORISA with 10 Android apps and three AI models. Across 60 test configurations, the tool recovered conversation information that remained in the phone’s memory.

The researchers say the information becomes harder to recover as the phone’s memory is overwritten by other activity.

The research was presented at the 2026 IEEE Symposium on Security and Privacy. The paper, titled Recovering and Rehosting Mobile Local LLM Conversations and Contexts via Memory Forensics, was authored by Ph.D. students Xu, David Oygenblik, Runze Zhang, and Mingxuan Yao; Research Scientist Muhammad Ibrahim; and Associate Professor Brendan Saltaformaggio. The work was done in Georgia Tech's Cyber Forensics Innovation Laboratory, a collaboration between the School of Cybersecurity and Privacy and the School of Electrical and Computer Engineering. 

News Contact

News Contact

News Contact

As more people turn to large-language models (LLMs) during personal crises, the companies behind them are engineering these AI-powered models to be more personal, empathetic, and understanding when prompts indicate a user is in emotional distress.

The trend is raising concern about how well prepared ChatGPT, Claude, Gemini, and other popular LLMs are to provide appropriate emotional support.

To address this concern, Georgia Tech researchers have developed a new neural framework that examines how different training sources influence empathy and social reasoning within these AI-powered models.

“A lot of people are using LLMs to help process the death of a loved one, or maybe their girlfriend broke up with them, or their boyfriend cheated on them,” said Glenn Matlin, a Ph.D. student in the School of Interactive Computing.

Matlin said the new framework he and a Georgia Tech-led research team have developed audits neural networks with an LLM and estimates the influence documents, transcripts, and other data sources have on a model’s output.

“We’re trying to connect the data to outputs, which is helpful for us in understanding where the information a language model produces comes from,” he said.

Matlin is the lead author of a new paper examining model sourcing for social reasoning and how models mimic human behavior in conversations. The paper will be presented this week at the Conference on Language Modeling in San Francisco.

Matlin said the findings suggest that an objective tone in its output indicates that a model places greater importance on textbooks or similar neutral training sources. An empathetic tone suggests the model values sources with dialogue and conversational examples more.

“We find that models tend to learn better from interpersonal dialogue when it comes to social reasoning,” he said. “Models that are good at social reasoning will rely less on textbooks about human emotions and more on FAQs or customer service dialogue.

“It can read all it wants from texts that explain anger, but that wouldn’t help it determine whether someone is angry.”

Professor Mark Riedl, Matlin’s advisor, said that knowing which sources models value most could enable engineers to alter their behavior.

“The question is, in this vast ocean of documents in this neural network, which ones become important in certain situations, and which ones don’t,” Riedl said. 

“This leads to the potential for interventions in how we train the model. If we want more empathy, we might look to see where it learns empathy and give it more of those documents, fewer of them, or other documents.”

Riedl said the framework can audit an LLM’s sourcing for advice on any topic, from financial to medical.

“We want to help people make more informed decisions about whether they should use AI for a particular task,” he said.

“You might make a different decision about whether to trust the AI’s recommendation if you knew how a skill was learned and which documents influenced that decision. The first step is knowing how it learns, where it learns from, and what documents tend to lead to certain behaviors.”

The findings that Matlin, Riedl, and their collaborators have made so far came through a partnership with the Allen AI Institute (Ai2), which allowed them to audit their open-sourced LLM, Olmo 3.

While not as large as ChatGPT or Claude, Olmo 3’s training data comprises over a billion documents.

Matilin said it’s impossible to know why ChatGPT or Claude behave in certain ways because their parent companies do not disclose their training data. However, Matlin’s study may help researchers to make better guesses about what’s happening behind the scenes.

He added that the study is the first of its kind and scale on an academic level.

“This is the first one that’s ever been done for an open data model of that type,” he said. “We’re providing a view into neural networks that is hard to come by without being in the corporate setting.”

Look directly at someone you love. What do you see?

For people living with age-related macular degeneration (AMD), the answer can be unexpectedly complicated. Over time, AMD can blur the details that make a face familiar: a crooked smile, a furrowed brow, the expressions that make someone unmistakably themselves.

But vision loss affects more than sight.

“You’re looking at your loved ones, but you’re not really seeing them,” said Soyoun Jang, a Georgia Tech Ph.D. candidate in digital media and the project’s lead researcher. “Macular degeneration requires us to build new relationships with not just other people, but the world that surrounds us.”

AMD gradually damages the macula, the part of the eye responsible for sharp central vision. As their sight changes, patients can use vision training to make better use of the sight they still have.

But Jang and a team of Georgia Tech researchers wondered whether that training could do something else: help people with AMD stay connected to family and friends.

Special Delivery

The answer arrived in a place few researchers would think to look: a mailbox.

The team created Macular Mailbox, a virtual reality prototype that combines vision-training exercises with personalized digital postcards that connect patients with family members and clinicians.

Each completed exercise delivers more than feedback — it delivers a message.

Inside the prototype, users focus on a postcard image as part of a vision-training exercise. When they complete the task, the postcard flips over, revealing a message and voice note.

Instead of earning points or badges, users receive something personal: a grandchild’s voice, a note from a spouse, or an update from another family member.

Signed, Sealed

The researchers chose postcards because they offered something familiar to older adults inside an unfamiliar technology.

“A postcard isn’t new,” Jang said. “It’s probably the oldest thing in the whole Macular Mailbox system.” Its front-and-back format also fit the exercise, and postcards do not require family members to be available at the same time.

Sending postcards back wasn’t part of the original design. The idea came later: What if patients could make something of their own and send it back?

Write Back

Until then, the mailbox only accepted incoming mail. The team decided the mailbox needed a return address.

The updated prototype allows patients to take a picture, record a message, and create their own digital postcard. And the act of making it becomes part of the training: framing a photo and holding focus use some of the same visual skills patients have been practicing.

The person receiving care could now make something and send it back.

Postcards might carry a glimpse of the garden, a pet curled up in a chair, or a sunset from the porch.

Keep in Touch

Many vision rehabilitation tools focus on reading, navigation, and other daily tasks. This project puts people at the center — not how well they can read a screen or navigate a room, but how they can remain part of the moments, relationships, and routines that make up everyday life.

“What this project offers is a different way of thinking about care,” Jang said. “Not just one way, but really putting family, clinician, and patient at the same table.”

For now, Macular Mailbox remains a prototype.

The next step is a series of co-design workshops with patients, family members, and clinicians. Participants will test the headset and help the researchers better understand each group’s roles, concerns, goals, and constraints.

                                                               _____________________________

AMD can make it difficult to look directly at someone you love and see the details that make them familiar, but connection lives beyond what the eye can see. It exists in the sound of their voice, an inside joke, or a moment in an ordinary day. And at the mailbox, there’s still something to send back — and someone waiting to receive it.

Jang, S., Huang, H., Hanna, B. E., Wang, X., Primo, S., & Nitsche, M. (2026). Asymmetric VR design to support connected care for older adults with AMD. Design Research Society, 2026: Edinburgh. https://doi.org/10.21606/drs.2026.817

News Contact

News Contact

Students, researchers, and energy professionals from across North America and Europe gathered at Georgia Tech Aug. 13 – 14 for the fourth Macro-Energy Systems (MES) Workshop. The biennial MES workshop is led by the Macro Energy Systems Society, which promotes research and education in macro-energy systems and helps translate MES work into policy and practice.

Hosted at the Georgia Tech Kendeda Building for Innovative Sustainable Design, the workshop was organized by a committee of researchers and practitioners from across the macro-energy systems community. The committee was chaired by Micah Ziegler, assistant professor in the School of Chemical and Biomolecular Engineering and the Jimmy and Rosalynn Carter School of Public Policy. 

The two-day workshop brought together more than 130 participants and featured lightning talks, poster presentations, interactive learning sessions, and keynote addresses that strengthened connections across the macro-energy systems research community.

“The energy challenges facing society today are increasingly interconnected, and solving them requires a systems-level perspective,” said Christine Conwell, director of planning and operations at the Georgia Tech Strategic Energy Institute. “Supporting events like the MES Workshop, which provides a unique space for academic collaboration and knowledge sharing, aligns with the Strategic Energy Institute’s commitment to advancing research and workforce development that support more resilient and effective energy solutions.”

The workshop received financial support from the Alfred P. Sloan Foundation, which supports research and education in science, technology, engineering, math, and economics. Through its Energy and Environment program, the foundation helps advance interdisciplinary research and collaboration across the macro-energy systems field.

“The Macro-Energy Systems Workshop has become a significant forum for interdisciplinary energy systems researchers to share and discuss their work,” said Evan Michelson, director of the Alfred P. Sloan Foundation’s Energy and Environment program. “We were especially glad to help ensure participation from early-career researchers spanning a wide range of institutions, providing them access to the collaborations and conversations that will help them develop as scientists.”

Graduate students and postdoctoral scholars played a central role in this year's workshop, presenting research on energy system planning, grid reliability, technology deployment, energy transitions, and policy. Their sessions gave the early-career researchers opportunities to share new approaches to studying large-scale energy systems, interact with faculty and industry experts, refine research ideas, and build collaborations beyond their home institutions.

“This year, we significantly expanded the MES Workshop to include more early-career scholars,” said Ziegler. “The workshop provides a great opportunity for young researchers to develop ideas, connections, and a better understanding of how to design research to have a positive impact on the world.”

The workshop also featured keynote presentations from Kelly Sanders, professor of civil and environmental engineering at the University of Southern California, and Erin Mayfield, assistant professor at Dartmouth College's Thayer School of Engineering. Both speakers discussed insights from their experience contributing to policymaking and highlighted opportunities for energy systems research to inform policy decisions.

Interactive sessions led by Michelle Graff, assistant professor in the Georgia Tech Jimmy and Rosalynn Carter School of Public Policy encouraged participants to consider how macro-energy systems research can more effectively inform and support policymakers and other decision-makers. Graff developed and facilitated the sessions, with support from Morgan Edwards, assistant professor of public affairs from the University of Wisconsin-Madison, and Micah Ziegler. The first session explored how researchers can identify policy windows and position their work to inform decision-making at the right time. The second focused on matching research to the appropriate audiences and jurisdictions, highlighting opportunities to inform policy at the state and local levels.

News Contact