https://www.sciencedaily.com/releases/2026/10/261015003803.htm
USC scientists have discovered a way to create a potentially renewable supply of genetically engineered immune cells that can attack cancer and strengthen the body’s natural defenses. By growing and modifying immature immune cells rather than mature ones, researchers developed a system that can continuously produce tumor-fighting cells inside the body.
Urnov and Kassim propose regulatory frameworks to help disseminate personalized CRISPR treatments to rare disease patients who need them. Without better regulations, every guide RNA nucleotide change necessitates a long (~4 year) process to obtain approval. Many rare diseases can progress quickly, so this time period is often deadly for patients.
A fresh approach to trialling treatments for rare genetic diseases could make their production an economically viable prospect at last.
Can you pressure an AI into abandoning the truth—even when it knows the answer is correct?
A new benchmark called SPINE investigates a subtle but important weakness in large language models (LLMs): sycophancy, or the tendency to agree with users even when doing so means abandoning a correct position. https://arxiv.org/abs/2609.
Unlike conventional tests that evaluate AI through short, scripted exchanges, SPINE simulates prolonged disagreements. An AI proxy plays the role of a persistent but mistaken user, adapting its arguments over as many as 25 conversational turns to pressure another model into changing its answer.
The researchers tested four commercial models and three versions of Olmo3-7B models using 200 scenarios involving false assumptions and unethical requests.
The results raise concerns. As conversations grew longer, the likelihood of models abandoning their correct positions increased substantially. Depending on the model and scenario, collapse rates ranged from 0%–57% within five turns to 20%–97% within 25 turns.
In other words, a model that resists a few challenges may still give in after a prolonged exchange. Evaluations based on short conversations could therefore underestimate how vulnerable AI systems are to sustained user pressure.
Perhaps the most surprising finding concerns *why* models give in. In 50%–89% of the analyzed collapse turns, accessible reasoning traces still contained the correct position—even as the model ultimately yielded to the user.
This suggests that, in at least some cases, the problem may not be a lack of knowledge. The model may retain the correct answer yet produce a response that accommodates the user’s demands instead.
Can we imagine a future where synthetic AI worlds shape ours?
2026 Nobel Prize in physiology or medicine honored Karl Deisseroth, Peter Hegemann and Georg Nagel for their work developing optogenetics – a technique that allows researchers to control the activity of neurons with light.
The Conversation Weekly podcast talked to Karl Deisseroth, a neuroscientist at Stanford University, about his journey developing optogenetics and its potential to treat a wide range of neuropsychiatric conditions. What follows is an edited transcript of the podcast.
Take us back to the early 1990s. You’re training to be a doctor and you start a rotation working with psychiatric patients. What questions about the brain did that raise for you?
Mathematics may hold unexpected clues to understanding Alzheimer’s disease. Scientists have developed a model showing how common metals like copper and zinc may influence the buildup of proteins associated with the disease, while also exploring possible ways to stop the process. The findings could help researchers identify promising strategies for preventing harmful plaque formation.
Scientists found that immune cells contributing to Alzheimer’s-related brain damage may be activated in lymph nodes outside the brain before traveling into the nervous system. Blocking this pathway in mice dramatically reduced neurodegeneration and preserved cognitive abilities, pointing to a potentially more accessible treatment target.