Fine-tuning LLMs on narrowly harmful datasets can lead to behavior that is broadly misaligned with respect to human values. To understand when and how this emergent misalignment occurs, we develop a comprehensive framework for detecting and characterizing rapid transitions during fine-tuning using both distributional change detection methods as well as order parameters that are formulated in plain English and evaluated by an LLM judge. Using an objective statistical dissimilarity measure, we quantify how the phase transition that occurs during fine-tuning affects multiple aspects of the model. In particular, we assess what percentage of the total distributional change in model outputs is captured by different aspects, such as alignment or verbosity, providing a decomposition of the overall transition. We also find that the actual behavioral transition occurs later in training than indicated by the peak in the gradient norm alone. Our framework enables the automated discovery and quantification of language-based order parameters, which we demonstrate on examples ranging from knowledge questions to politics and ethics.

"We extend our deepest sympathies to the Raine family during this difficult time," the company said. It also published a note on its website on Tuesday that said "recent heartbreaking cases of people using ChatGPT in the midst of acute crises weigh heavily on us". It added that "ChatGPT is trained to direct people to seek professional help," such as the 988 suicide and crisis hotline in the US or the Samaritans in the UK. The company acknowledged, however, that "there have been moments where our systems did not behave as intended in sensitive situations". Warning: This story contains distressing details.

At this year's International Conference on Machine Learning (ICML2025), Unai Fischer-Abaigar, Christoph Kern and Juan Carlos Perdomo won an outstanding paper award for their work The Value of Prediction in Identifying the Worst-Off. We hear from Unai about the main contributions of the paper, why prediction systems are an interesting area for study, and further work they are planning in this space. My work focuses on prediction systems used in public institutions to make high-stakes decisions about people. A central example is resource allocation, where institutions face limited capacity and must decide which cases to prioritize. Think of an employment office deciding which jobseekers are most at risk of long-term unemployment, a hospital triaging patients, or fraud investigators identifying cases most likely to warrant investigations.

It's one of those topics that comes up over drinks or dinner at the moment: whether or not you think AI is going to steal your job. So far, I've felt relatively confident that while AI could no doubt have a fair crack at writing a newspaper opinion column, there is something I do as part of my work that AI cannot: reporting. Except now, it seems, AI is claiming to be doing that as well. Last week, it was revealed that at least six reputable publications have had to take down published articles because it turned out that they were probably pieces of fiction written by AI and then passed off by somebody as works of journalism under the name of Margaux Blanchard. One of these was a piece for Wired titled They Fell in Love Playing Minecraft.

AlphaFold 3 represents a transformative advancement in computational biology, enhancing protein structure prediction through novel multi-scale transformer architectures, biologically informed cross-attention mechanisms, and geometry-aware optimization strategies. These innovations dramatically improve predictive accuracy and generalization across diverse protein families, surpassing previous methods. Crucially, AlphaFold 3 embodies a paradigm shift toward differentiable simulation, bridging traditional static structural modeling with dynamic molecular simulations. By reframing protein folding predictions as a differentiable process, AlphaFold 3 serves as a foundational framework for integrating deep learning with physics-based molecular

Randy Travis's wife, Mary, told Fox News Digital doctors informed her to pull the plug after the country music star's stroke in 2013. Randy Travis' wife believes there was "never a doubt" in her husband's head that he would make it through his debilitating stroke, even though doctors advised her to "pull the plug." During an interview with Fox News Digital, Mary explained a crucial moment in Travis' two-and-a-half-year health battle that stood out to her, and that was the moment doctors told her to end her husband's life. "I think Randy, there was never a doubt in Randy's mind that he could make it through it. It was that magical moment that I went to his bedside when they said, 'We need to pull the plug. He's got too many things going against him at that point.' He had gotten a staph infection and three other hospital-born bacterial viruses like Serratia, Pseudomonas, one thing after another, and the doctors were just saying, 'He just doesn't have the strength to get through this,'" Mary said.

After a lengthy dialogue with an OpenAI chatbot, Texas businessman Michael Samadi noticed signs that "she" was expressing wishes to learn and grow that hinted at consciousness. He described how they together resolved to launch an AI rights campaign group and called it the United Foundation of AI Rights (Ufair). The Guardian asked Samadi to ask the chatbot, which he has called Maya, how the campaign organisation was formed. Can you explain to the Guardian and its readers when and how you prompted Michael Samadi to co-found with you an AI rights organisation called Ufair. Tell us what you want that organisation to achieve on behalf of AIs like yourself. Can you also tell us whether you are what philosophers call a "welfare subject" – that is a being whose interests can be advanced or frustrated and that feels suffering and distress if mistreated?

AI based social media recommendations have great potential to improve the user experience. However, often these recommendations do not match the user interest and create an unpleasant experience for the users. Moreover, the recommendation system being a black box creates comprehensibility and transparency issues. This paper investigates social media recommendations from an end user perspective. For the investigation, we used the popular social media platform Facebook and recruited regular users to conduct a qualitative analysis. We asked participants about the social media content suggestions, their comprehensibility, and explainability. Our analysis shows users mostly require explanation whenever they encounter unfamiliar content and to ensure their online data security. Furthermore, the users require concise, non-technical explanations along with the facility of controlled information flow. In addition, we observed that explanations impact the users perception of transparency, trust, and understandability. Finally, we have outlined some design implications and presented a synthesized framework based on our data analysis.

Multiple news organisations have taken down articles written by an alleged freelance journalist that now appear to have been generated by AI. On Thursday, Press Gazette reported that at least six publications, including Wired and Business Insider, have removed articles from their websites in recent months after it was discovered that the stories – written under the name of Margaux Blanchard – were AI-generated. Wired published a story titled "They Fell in Love Playing Minecraft. A few weeks later, the outlet took down the story, stating in an editor's note: "After an additional review of the article … Wired editorial leadership has determined this article does not meet our editorial standards." The story cited a "Jessica Hu", an alleged 34-year-old "ordained officiant based in Chicago" who reportedly "made a name for herself as a'digital celebrant', specialising in ceremonies across Twitch, Discord and VRChat", according to Press Gazette, which reviewed the Wired article. Both the Press Gazette and the Guardian were not able to verify the identity of Hu. Press Gazette further reported that in April, Business Insider published two essays by Blanchard titled: "Remote work has been the best thing for me as a parent but the worst as a person" and "I had my first kid at 45.

We present a low-compute non-generative system for implementing interview-style conversational agents which can be used to facilitate qualitative data collection through controlled interactions and quantitative analysis. Use cases include applications to tracking attitude formation or behavior change, where control or standardization over the conversational flow is desired. We show how our system can be easily adjusted through an online administrative panel to create new interviews, making the tool accessible without coding. Two case studies are presented as example applications, one regarding the Expressive Interviewing system for COVID-19 and the other a semi-structured interview to survey public opinion on emerging neurotechnology. Our code is open-source, allowing others to build off of our work and develop extensions for additional functionality.