Character.AI bans users under 18 after being sued over child's suicide Move comes as lawmakers move to bar minors from using AI companions and require companies to verify users' age The chatbot company Character.AI will ban users 18 and under from conversing with its virtual companions beginning in late November after months of legal scrutiny. The announced change comes after the company, which enables its users to create characters with which they can have open-ended conversations, faced tough questions over how these AI companions can affect teen and general mental health, including a lawsuit over a child's suicide and a proposed bill that would ban minors from conversing with AI companions. "We're making these changes to our under-18 platform in light of the evolving landscape around AI and teens," the company wrote in its announcement. "We have seen recent news reports raising questions, and have received questions from regulators, about the content teens may encounter when chatting with AI and about how open-ended AI chat in general might affect teens, even when content controls work perfectly." Last year, the company was sued by the family of 14-year-old Sewell Setzer III, who took his own life after allegedly developing an emotional attachment to a character he created on Character.AI.

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 attend AIES and ECAI, learn about policy design for two-sided platforms, discover how to balance speed and physical laws in atomic-scale simulations, and find out more about machine learning for chip design. October has been a busy month on the conference front. Over in Madrid, researchers gathered for the conference on Artificial Intelligence, Ethics, and Society (AIES) . The event featured two keynote talks, panel discussions and poster sessions.

Large Language Model (LLM)-based agents have emerged as a transformative approach for open-ended problem solving, with information seeking (IS) being a core capability that enables autonomous reasoning and decision-making. While prior research has largely focused on improving retrieval depth, we observe that current IS agents often suffer from low search efficiency, which in turn constrains overall performance. A key factor underlying this inefficiency is the sparsity of target entities in training tasks, which limits opportunities for agents to learn and generalize efficient search behaviors. To address these challenges, we propose WebLeaper, a framework for constructing high-coverage IS tasks and generating efficient solution trajectories. We formulate IS as a tree-structured reasoning problem, enabling a substantially larger set of target entities to be embedded within a constrained context. Leveraging curated Wikipedia tables, we propose three variants for synthesizing IS tasks, Basic, Union, and Reverse-Union, to systematically increase both IS efficiency and efficacy. Finally, we curate training trajectories by retaining only those that are simultaneously accurate and efficient, ensuring that the model is optimized for both correctness and search performance. Extensive experiments on both basic and comprehensive settings, conducted on five IS benchmarks, BrowserComp, GAIA, xbench-DeepSearch, WideSearch, and Seal-0, demonstrate that our method consistently achieves improvements in both effectiveness and efficiency over strong baselines.

LLM-based search agents are increasingly trained on entity-centric synthetic data to solve complex, knowledge-intensive tasks. However, prevailing training methods like Group Relative Policy Optimization (GRPO) discard this rich entity information, relying instead on sparse, outcome-based rewards. This critical limitation renders them unable to distinguish informative "near-miss" samples-those with substantially correct reasoning but a flawed final answer-from complete failures, thus discarding valuable learning signals. We address this by leveraging the very entities discarded during training. Our empirical analysis reveals a strong positive correlation between the number of ground-truth entities identified during an agent's reasoning process and final answer accuracy. Building on this insight, we introduce Entity-aware Group Relative Policy Optimization (E-GRPO), a novel framework that formulates a dense entity-aware reward function. E-GRPO assigns partial rewards to incorrect samples proportional to their entity match rate, enabling the model to effectively learn from these "near-misses". Experiments on diverse question-answering (QA) and deep research benchmarks show that E-GRPO consistently and significantly outperforms the GRPO baseline. Furthermore, our analysis reveals that E-GRPO not only achieves superior accuracy but also induces more efficient reasoning policies that require fewer tool calls, demonstrating a more effective and sample-efficient approach to aligning search agents.

Unlike other military technologies driven by national security needs and developed with federal funding, AI is predominantly funded and advanced by commercial industry for civilian applications. However, there is a lack of understanding of the reasons commercial AI firms decide to work with the DoD or choose to abstain from the defence market. This thesis argues that the contract law and procurement framework are among the most significant obstacles. This research indicates that the commercial AI industry actually views the DoD as an attractive customer. However, this attraction is despite the obstacles presented by traditional contract law and procurement practices used to solicit and award contracts. Drawing on social exchange theory, this thesis introduces a theoretical framework, optimal buyer theory, to understand the factors that influence a commercial decision to engage with the DoD. Interviews from a sample of the participants explain why the AI industry holds such perceptions, opinions, and preferences about contracts generally and the DoD, specifically, in its role as a customer. This thesis concludes that commercial AI firms are attracted to contracts that are consistent with their business and technology considerations. Additionally, it develops best practices for leveraging existing contract law, primarily other transaction authority, to align contracting practices with commercial preferences and the machine learning development and deployment lifecycle.

Prunella Scales, who died at the age of 93, was one of Britain's finest comic actors. But despite a long and distinguished career on stage and screen, she will inevitably be remembered as Sybil Fawlty in the 1970s TV comedy, Fawlty Towers. It was Sybil's mission in life to keep tabs on her stick insect husband Basil - played by John Cleese - between cigarette-fuelled phone conversations with her friend, Audrey. It fell to her to placate guests who had been shouted at, totally ignored or, in some cases, throttled by Basil when in one of his more manic moods. Her nightmarish laugh, gravity-defying hairdo and ferocious temper were part of a carefully constructed character that ranks as a comic masterpiece.

The Atlantic CEO's new book,, examines his complicated relationship with the sport. On this week's episode of, he talks about the ways tech is helping him become a better runner. To most of the world, Nicholas Thompson is known as an editor, an AI enthusiast, or something of a LinkedIn influencer. But the former WIRED editor in chief, who is now CEO of The Atlantic, is often better known to colleagues as . On Tuesday, Thompson is releasing . As the title suggests, it's a book about his commitment to running--Thompson runs a ridiculously fast marathon and holds the American 50K record for the 45-49 age group. Ultimately, though, the book examines the complicated relationship between the sport, Thompson, and his father, who first took him on a run when he was just 5 years old. Tech obsessives, of course, will also get their fix: includes plenty of science-backed training guidance and documents Thompson's experience training with elite Nike coaches. On this week's episode of, I talked to Thompson (who was also my first boss; he hired me as an intern at WIRED in 2008) about his book, the interplay between running and addiction, and what he thinks AI can do for runners for writers. It is a joy to be here with you at Condé Nast at WIRED. I loved coming up those elevators. I love seeing you as the editor in chief. I'm thrilled that you're here. We're going to start this conversation the way we start all of them, which is with a little warmup, some rapid-fire questions. In honor of your new book,, I'm gonna make them entirely running themed. I mean, if your listeners don't wanna hear about running Trail run or track run? Worst running injury you've ever had. The one you wish people would stop talking to you about. You only need to run a 20-miler before a marathon. What do you need to run? Why do people die at mile 20? Because they only train for [marathons] with 20-mile-runs. I generally prefer people, but then you have to schedule it. Backup sport of choice if you could never run again.

'Every user has to make a choice of actually wanting to download Firefox and use it.' 'Every user has to make a choice of actually wanting to download Firefox and use it.' Do you need an assistant for your online activities? Multiple major players in artificial intelligence are moving on from chatbots like ChatGPT and are now focusing their efforts on new browsers with deep AI integrations. Those could take the form of an agent that shops for you or an omnipresent chatbot that follows you around and summarizes what you're seeing, looks up related stuff, or answers related questions.

The 28th European Conference on Artificial Intelligence (ECAI-2025) is currently taking place in Bologna, Italy, running from 25-30 October 2025. During the opening ceremony, the winners of the ECAI-2025 and Prestigious Applications of Intelligent Systems (PAIS-2025) outstanding paper awards were announced. Letting AI agents interact in multi-agent applications adds a layer of complexity to the interpretability and prediction of AI outcomes, with profound implications for their trustworthy adoption in research and society. Game theory offers powerful models to capture and interpret strategic interaction among agents, but requires the support of reproducible, standardized and user-friendly IT frameworks to enable comparison and interpretation of results. We describe its implementation and usage, and we employ it to uncover biased outcomes in popular games among AI agents, depending on the employed Large Language Model (LLM) and used language, as well as on the personality trait or strategic knowledge of the agents.

How can large language models (LLMs) serve users with varying preferences that may conflict across cultural, political, or other dimensions? To advance this challenge, this paper establishes four key results. First, we demonstrate, through a large-scale multilingual human study with representative samples from five countries (N=15,000), that humans exhibit significantly more variation in preferences than the responses of 21 state-of-the-art LLMs. Second, we show that existing methods for preference dataset collection are insufficient for learning the diversity of human preferences even along two of the most salient dimensions of variability in global values, due to the underlying homogeneity of candidate responses. Third, we argue that this motivates the need for negatively-correlated sampling when generating candidate sets, and we show that simple prompt-based techniques for doing so significantly enhance the performance of alignment methods in learning heterogeneous preferences. Fourth, based on this novel candidate sampling approach, we collect and open-source Community Alignment, the largest and most representative multilingual and multi-turn preference dataset to date, featuring almost 200,000 comparisons from annotators spanning five countries. We hope that the Community Alignment dataset will be a valuable resource for improving the effectiveness of LLMs for a diverse global population.