The One Hundred Year Study on Artificial Intelligence (AI100) is a longitudinal study that assesses AI's development and societal impact. Reports are released once every five years, with the 2026 edition being the third in the series. The title of this year's report is . The panel behind the project is an international committee of experts tasked with examining AI's evolution across technical, ethical, economic, and societal domains. They respond to 12 standing questions about AI's advances, opportunities, dangers, and societal effects, and compare developments since 2021 with the previous panel's assessments.

Today, most of us interact with AI assistants - reactive bots that wait for human instructions. Yet, AI assistants are rapidly being replaced by agentic AI agents, that can interact with external tools, browse the web, generate images, send emails, and perform increasingly complex workflows on behalf of users. As these agentic agents become more capable, they may be exploited by people with bad intentions. Safety tests have generally only checked whether an AI agent refuses a single harmful request - they don't measure what happens when an attacker gradually persuades the AI through a series of seemingly harmless conversations. Now, researchers from EPFL's Natural Language Processing Laboratory have developed STING (Sequential Testing of Illicit N-step Goal execution), an automated testing framework that simulates how a real attacker might manipulate large language model (LLM) agents into carrying out harmful tasks over multiple interactions.

Minji Park is a PhD candidate at the Korea Institute of Energy Technology (KENTECH). In her paper, with co-authors Seunghyun Yoon and Hyuk Lim, she introduces a framework which aims to help make large language models stay more up-to-date when pieces of information change. We hear from Minji about inspired her work, how the framework operates, and her plans to further develop this research. What is the topic of the research in your paper and why is it an interesting area for study? The paper is fun, because we are challenging the norm.

Cornell researchers are using artificial intelligence to speed the search for better battery materials while using a creative approach to chemistry to expand the design space for electrolytes, unlocking new possibilities for safer, higher-performing energy storage. Electrolytes shuttle ions between battery electrodes and are key to a battery's energy density, lifetime and charging speed, yet they are difficult to design because small changes in composition can dramatically change how they work. "Electrolytes are not just supporting materials in batteries; they are a design space," said Fengqi You . You's research group integrates theoretical frameworks, artificial intelligence and other computational methods to help discover next-generation battery materials, among other research themes. His group has introduced IonNet, an AI framework that predicts how well lithium ions move through solid materials using only their chemical composition.

"With AI, we need to reconsider what it means to learn" Generative AI has worked its way into every field over the past four years and is prompting us to reconsider our habits. At universities, students are using it to write, find information and solve problems, while scientists have already incorporated it largely into their research. Teachers are being called on to rethink how they do their job now that AI programs can complete - in just seconds - a number of tasks traditionally assigned to students. These are just some of the questions we explore in our special online report on AI at universities, published at the start of this new school year . These are also questions that Schwartz - a professor of educational technology at the Stanford Graduate School of Education and a faculty director at the Stanford Accelerator for Learning - has been studying for the past few years.

This post contains a list of the AI-related seminars that are scheduled to take place in the next couple of months. All events detailed here are free and open for anyone to attend virtually. Stephanie Gil (Harvard University) EPFL The Zoom link is here . Timo Bertholdt and Imre Polik (FICO) Association of European Operational Research Societies To receive the seminar link, sign up to the mailing list . Jonah Botvinick-Greenhouse (Yale University) University of Minnesota Zoom registration is here .

Welcome to our monthly digest, where you can catch up with any AIhub stories you may have missed, peruse the latest news, recap recent events, and more. This month, we chat to AI pioneer, Michael Littman, learn about time-series forecasting for clinical data, investigate evolutionary game theory for cybersecurity, and get the scoop on neuromorphic computing. In the latest interview in our series meeting AI pioneers, Michael Littman reflects on the evolution of reinforcement learning, the power of science communication, and how generative AI is changing the way we teach and learn. Mayra Elwes is a PhD student developing machine learning methods for clinical time-series data. She is working on practical strategies to evaluate methods in a more contextualised way and on improving existing domain adaptation methods specific to time-series forecasting.

Systems that perform well in controlled lab settings often fall short when deployed in homes, in hospitals or on city streets. Vision-based systems, for example, hold promise for improving safety, monitoring health and aiding rehabilitation. For example, a camera-based system could help a clinician assess a patient's gait or rehabilitation progress. In the future, an assistive robot in an older adult's home might need to recognize when the person is struggling to stand or reaching for an object. But a major challenge is ensuring these robots perform reliably beyond the lab, particularly in messy, unpredictable or low-visibility conditions. As a computer scientist who studies AI systems designed to understand human behavior - especially through visual cues like body movements and interactions with objects - I see three main challenges to taking AI from the lab to the real world: generalization, human behavior and resources.

Researchers get AI'drunk' to expose new cyber security risks in chatbots Fine-tuning large language models using drunken text makes them more likely to leak secrets and answer harmful questions, new UNSW-led research shows. UNSW researchers have shown that large language models (LLMs) can be pushed to imitate drunken behaviour - and that once they do, they are significantly more likely to leak confidential information and answer questions they are designed to refuse. In the study, conceptualised and led by Dr Aditya Joshi, Senior Lecturer, School of Computer Science and Engineering, with Anudeex Shetty (Research Assistant) and Professor Salil Kanhere, researchers from UNSW's School of Computer Science and Engineering (CSE) found that AI trained or prompted to mimic drunken speech patterns became more vulnerable to jailbreaking and privacy breaches than their'sober' counterparts. The UNSW team tested three methods of inducing'drunk' behaviour in LLMs. The vulnerability wasn't tested across every LLM on the market.

AI is creating jobs as well as erasing them - but how rewarding are they? Don't believe the hype about AI taking over all forms of work . Most of the evidence to date suggests large-scale job losses are limited . But what is changing is the way work is organised. One area of increased employment is in people "cleaning up" after large language models (LLMs), or quality controlling them.