News
Helping Lay the Mathematical Foundation for More Trustworthy AI
UMD’s Soheil Feizi is part of a $7.5M multi-institutional initiative to advance the next generation of artificial intelligence.
TRAILS Fellows Map a More Trustworthy Path for AI
A team of five UMD undergraduates developed an interactive tool to embed ethics and public trust in AI research.
AI Can Enforce the Rules. Can It Build an Online Community?
UMD researchers identify where language models could help volunteer moderators—and where human judgment remains essential.
Evaluating AI Systems Requires More Than Technical Metrics
TRAILS Postdoc Laura Fichtner emphasizes the need for participatory approaches that incorporate diverse viewpoints when assessing AI systems, alongside emerging efforts to develop national standards.
How We Talk About AI Reveals Whether We Trust It
UMD researchers will analyze everyday conversations for signs of trust—and distrust—in AI.
What Tech Workers Tell Their Kids About Careers in the Age of AI
UMD’s Katie Shilton studies how tech-worker parents are preparing their children for an AI-transformed workplace.
GW Researchers Receive New Grants to Study Applications of Trustworthy AI
Reporting from GW’s student newspaper highlights latest round of TRAILS seed grants.
When AI Writes for You, Can It Still Sound Like You?
UMD team finds that editing AI-generated text may not erase its stylistic fingerprints.
TRAILS Awards More Than $515K for Trustworthy AI Research
The fourth round of seed funding supports multidisciplinary teams tackling one of AI's biggest challenges: earning public trust.
New Tool Reveals Hidden Security Risks in AI Systems
UMD’s Michelle Mazurek warns that even safe AI models can be compromised by security vulnerabilities in the software infrastructure used to serve them.
AI Hiring Tools May Favor Their Own Work, UMD Study Finds
New research by UMD’s Jiannan Xu highlights emerging bias in AI-to-AI interactions—and practical fixes.
Team HAX Lab Wins National AI Innovation Challenge
Guided by Morgan State’s Naja Mack, the team earned top honors for a privacy-focused AI application designed to support communication and social well-being.
Can AI and People Play Nice?
UMD’s Jordan Boyd-Graber is uncovering what makes human-AI partnerships succeed through high-stakes quizbowl competitions, finding that trust may matter more than intelligence.
UMD Team Develops Precise “Undo Button” for AI Memory
UMD’s Soheil Feizi helped develop a new method removes problematic information from large language models without damaging their reasoning abilities.
The Role of Language in AI Research
UMD’s Marine Carpuat explains how multilingual AI systems can facilitate cross-language communication.
When the Feed Becomes the Front Line
UMD’s Cody Buntain is developing AI tools to analyze social media imagery from global conflicts, helping analysts establish ground truth while investigating the reliability of human-AI decision-making.
Rethinking the “Charisma Trap” in Human-AI Interaction
UMD’s Michelle Mazurek and GW’s Adam Aviv found that a chatbot’s polite demeanor creates a “charisma trap” that masks subtle biases and safety risks.
UMD Framework Used to Test Safety of Meta’s New Multimodal AI Model
UMD’s Furong Huang is co-leading efforts involving AI safety testing under simulated operational pressure.
How Generative AI Works—and Why It Can Mislead Users
GW’s David Broniatowski explains the pattern-based systems behind tools like ChatGPT, and the risks that come with them.
TRAILS Researchers Advance Robotics to Perform Complex Household Tasks
UMD’s Furong Huang and Tom Goldstein are advancing breakthroughs in trustworthy machine learning to create robotic systems that will excel in dynamic home environments.