Top 10 AI News
AI system learns from many types of scientific information and runs experiments to discover new materials
Published: September 25, 2025
The new “CRESt” platform could help find solutions to real-world energy problems that have historically challenged the materials science community ([news.mit.edu](https://news.mit.edu/2025/ai-system-learns-many-types-scientific-information-and-runs-experiments-discovering-new-materials-0925#:~:text=The%20new%20%E2%80%9CCRESt%E2%80%9D%20platform%20could,and%20engineering%20community%20for%20decades)). By integrating diverse scientific data and automated experiments, this AI-driven system accelerates the discovery of new materials. Read full story.
New AI system could accelerate clinical research
Author: Adam Zewe | MIT News Published: September 25, 2025
By enabling rapid annotation of areas of interest in medical images, this AI tool can help scientists study new treatments and track disease progression ([news.mit.edu](https://news.mit.edu/2025/new-ai-system-could-accelerate-clinical-research-0925#:~:text=By%20enabling%20rapid%20annotation%20of,treatments%20or%20map%20disease%20progression)). In other words, it speeds up clinical research by quickly highlighting relevant image regions that researchers need to analyze. Read full story.
Improving the workplace of the future
Published: September 24, 2025
Economics PhD student Whitney Zhang investigates how new technologies and workplace policies affect labor markets. Her research explores how factors like scheduling practices impact employees’ well-being and productivity. This work offers insights into how AI and business decisions can shape the future of work ([news.mit.edu](https://news.mit.edu/2025/improving-workplace-future-whitney-zhang-0924#:~:text=Economics%20doctoral%20student%20Whitney%20Zhang,investigates%20how%20technologies%20and)). Read full story.
MIT affiliates win AI for Math grants to accelerate mathematical discovery
Author: Sandi Miller | Department of Mathematics Published: September 22, 2025
MIT math researchers David Roe and Andrew Sutherland have both received new AI for Math grants to advance automated theorem proving and other AI-driven mathematical tools. Their projects, along with those of additional MIT alumni awardees, aim to use AI to speed up and deepen mathematical research. This initiative could transform how mathematicians discover and prove new results in the future ([news.mit.edu](https://news.mit.edu/2025/ai-for-math-grants-accelerate-mathematical-discovery-0922#:~:text=Department%20of%20Mathematics%20researchers%20David,and%20Andrew%20Sutherland%20seek%20to)). Read full story.
New tool makes generative AI models more likely to create breakthrough materials
Author: Zach Winn | MIT News Published: September 22, 2025
Researchers have introduced SCIGEN, a tool that imposes structural constraints on generative AI models. With SCIGEN, scientists can steer AI generators to obey physical and chemical rules, helping to create novel materials with exotic properties (e.g., for quantum computing) ([news.mit.edu](https://news.mit.edu/2025/new-tool-makes-generative-ai-models-likely-create-breakthrough-materials-0922#:~:text=With%20SCIGEN%2C%20researchers%20can%20steer,for%20applications%20like%20quantum%20computing)). This advancement makes it more likely that AI will propose feasible breakthrough materials. Read full story.
How are MIT entrepreneurs using AI?
Author: Zach Winn | MIT News Published: September 22, 2025
At MIT’s delta v accelerator demo day, numerous student startups showcased how they integrate AI into products and processes. From logistics to health-tech, this year’s cohort demonstrated that AI is reshaping the startup landscape by enabling new solutions and efficiencies ([news.mit.edu](https://news.mit.edu/2025/how-are-mit-entrepreneurs-using-ai-0922#:~:text=This%20year%E2%80%99s%20delta%20v%20summer,process%20of%20building%20a%20startup)). Read full story.
What does the future hold for generative AI?
Author: Adam Zewe | MIT News Published: September 19, 2025
MIT convened its first Generative AI Impact Consortium Symposium, where researchers and industry leaders discussed anticipated breakthroughs and challenges in generative AI ([news.mit.edu](https://news.mit.edu/2025/what-does-future-hold-generative-ai-0919#:~:text=At%20the%20inaugural%20MIT%20Generative,advancements%20centered%20on%20this%20powerful)). Topics included how AI will evolve and what ethical, societal, and technical issues it may raise. Read full story.
How to build AI scaling laws for efficient LLM training and budget maximization
Author: Lauren Hinkel | MIT-IBM Watson AI Lab Published: September 16, 2025
Researchers at the MIT-IBM Watson AI Lab have published a “universal guide” for estimating how large language models will perform by extrapolating from smaller models ([news.mit.edu](https://news.mit.edu/2025/how-build-ai-scaling-laws-efficient-llm-training-budget-maximization-0916#:~:text=MIT,on%20smaller%20models%20in%20the)). By applying these scaling laws, teams can predict the accuracy and compute requirements of bigger models, and thus optimize their training budgets more efficiently. Read full story.
Regulators struggle to keep up with AI therapy apps
Published: September 30, 2025
As AI-driven mental health chatbots become popular amid care shortages, state and federal regulators are scrambling to create new rules. U.S. policymakers in several states have started imposing laws on AI therapy tools, but experts warn the patchwork of regulations is falling behind the rapid pace of technology innovation ([apnews.com](https://apnews.com/article/dfc5906b36fdd1fe8e8dbdb4970a45a7#:~:text=As%20AI%20therapy%20apps%20grow,inconsistent%20and%20often%20vague%2C%20leaving)). Read full story.
How Walmart plans to prepare America’s largest private workforce for an AI-driven future
Published: September 28, 2025
Walmart is proactively training and upskilling its vast workforce to adapt to an AI-driven labor market. The company launched a “Skills-First Workforce Initiative” emphasizing ability over credentials, aiming to help employees learn AI-related tasks and to promote a culture where workers can use new technology tools safely and effectively ([apnews.com](https://apnews.com/article/154ece8ba303ce6ac8c5030e6f719aa1#:~:text=Walmart%20is%20proactively%20preparing%20its,Workforce%20Initiative%20to%20help%20develop)). Read full story.
Microsoft reduces Israel’s access to cloud and AI products over reports of mass surveillance in Gaza
Published: September 25, 2025
Microsoft has restricted certain cloud and AI services for an Israeli military unit after an internal review found the unit’s use of Microsoft Azure was linked to mass surveillance of Palestinians ([apnews.com](https://apnews.com/article/3f4bf8036e7e02f385b258bd353af1fd#:~:text=Microsoft%20has%20restricted%20access%20to,This%20decision%20follows)). This decision follows investigative reports and represents a rare move by a tech company to halt access to its AI tools due to ethical concerns. Read full story.
Judge approves $1.5 billion copyright settlement between AI company Anthropic and authors
Published: September 25, 2025
A federal judge in San Francisco granted preliminary approval to a $1.5 billion settlement in a major copyright case. The lawsuit was brought by a group of authors and publishers who claimed the AI startup Anthropic had used about 465,000 books without permission to train its chatbot (Claude) ([apnews.com](https://apnews.com/article/9643064e847a5e88ef6ee8b620b3a44c#:~:text=A%20federal%20judge%20in%20San,465%2C000%20books%20to%20train%20its)). The settlement resolves one of the first big litigations over AI training data. Read full story.
Al Gore’s satellite and AI system is now tracking sources of deadly soot pollution
Published: September 24, 2025
Former Vice President Al Gore’s Climate TRACE initiative has expanded its satellite-and-AI tracking system to map soot (fine particulate) emissions worldwide. It now monitors thousands of pollution sources in over 2,500 cities using 300 satellites and 30,000 ground sensors, aiming to identify “super-emitter” sites of deadly air pollution ([apnews.com](https://apnews.com/article/8ba88e0b46f3ebdc35608d78f005de2f#:~:text=Former%20U,system%20aims%20to%20make%20data)). Read full story.
AI’s double-edged sword: UN leaders weigh its promise and peril
Published: September 24, 2025
At the UN Security Council meeting (Sept. 24, 2025), world leaders debated AI’s benefits and risks. They recognized AI’s transformative potential for peace and progress but also warned of its dangers, such as misinformation and autonomous weapons. The discussions emphasized the need for international cooperation to manage AI responsibly ([apnews.com](https://apnews.com/article/5cb0f21feeb3e734cc0c25eab67fe26d#:~:text=At%20a%20U,N)). Read full story.
OpenAI shows off Stargate AI data center in Texas and plans 5 more elsewhere with Oracle, Softbank
Published: September 23, 2025
OpenAI unveiled its new “Stargate” data center in Abilene, Texas – a joint project with Oracle and SoftBank. The massive facility, consisting of eight interconnected buildings, is designed to be the world’s largest AI “supercluster.” OpenAI plans to build five more such centers globally to support its growing AI computing needs ([apnews.com](https://apnews.com/article/0b3f4fa6e8d8141b4c143e3e7f41aba1#:~:text=OpenAI%20has%20unveiled%20its%20massive,The%20site%2C%20featuring%20eight)). Read full story.
Albania’s AI ‘minister’ makes its debut with an address to parliament
Published: September 18, 2025
Albania has introduced “Diella,” an AI-generated virtual minister, to its government. In a parliamentary address on Sept. 18, 2025, Prime Minister Edi Rama presented Diella (meaning “Sun” in Albanian) as a digital minister in traditional attire. Diella’s role is symbolic — aimed at showcasing Albania’s commitment to innovation and transparency ([apnews.com](https://apnews.com/article/3aa58c801d69b5b295975cc68079a2d3#:~:text=On%20September%2018%2C%202025%2C%20Albania,in%20traditional%20Albanian%20attire%20%E2%80%94)). Read full story.
Spotify Tightens AI Policy And Trims Catalog
Author: Bill Rosenblatt Published: September 26, 2025
Spotify announced new policies to clamp down on AI-generated music, spam, impersonation and fraud. The company reported removing over 75 million “spammy” tracks from its library as it implements stricter AI checks and content filters. This move is intended to ensure the quality and authenticity of music on the platform, in light of growing use of AI to produce music ([news.mit.edu](https://news.mit.edu/2025/ai-system-learns-many-types-scientific-information-and-runs-experiments-discovering-new-materials-0925#:~:text=The%20new%20%E2%80%9CCRESt%E2%80%9D%20platform%20could,and%20engineering%20community%20for%20decades)). Read full story.
Tesla’s Full-Self Driving Software Is A Mess. Should It Be Legal?
Author: Alan Ohnsman Published: September 23, 2025
Tesla’s full-self driving (FSD) system has been heavily promoted, but recent independent tests showed critical failures, such as ignoring stop signs and hitting obstacles. The debate centers on whether selling this yet-imperfect AI driving suite is safe or even legal, considering these flaws. Elon Musk insists that wider adoption is needed for improvement, making this an ongoing controversy in auto AI safety. Read full story.
Hidden Google Photos Code Suggests Return Of Missing Search Feature
Author: Paul Monckton Published: September 28, 2025
Code analysis of the Google Photos app indicates that Google may be reintroducing the “More Like This” search feature for images. If enabled, this AI feature would let users find photos that are visually similar, making the photo search experience more powerful. Google is experimenting with AI-driven photo search improvements to better help users organize and navigate their photo libraries. Read full story.
The 5 Technology Trends For 2026 Everyone Must Prepare For Now
Author: Bernard Marr Published: September 29, 2025
According to experts, five major technology trends will dominate in 2026: generalized AI assistants, quantum computing advances, new energy storage/production, intelligent agents in everyday tasks, and a renewed focus on human skills alongside machines. Companies and professionals are advised to start preparing now for these shifts, as they will drive innovation and reshape industries by next year. Read full story.
AI Safety Group Finds Agentic Tools Slow Down Open Source Devs
Author: John Werner Published: September 28, 2025
According to a study by the MIT-based METR project, giving experienced open-source developers advanced AI coding assistants actually slowed them down by an average of 19%. The findings highlight issues like the learning curve and integration challenges of new AI “agentic” tools. The result cautions against overestimating current AI productivity gains in software development and stresses careful benchmarking and training. Read full story.
Why AI Evals And KPIs Are The New Standard For Scaling Healthcare AI
Author: Sahar Hashmi Published: September 28, 2025
In healthcare, establishing clear evaluation metrics and KPIs (key performance indicators) is becoming essential as AI tools are adopted. By rigorously measuring an AI application’s accuracy, reliability, and clinical impact, hospitals can ensure new AI tools truly improve patient care safely. This data-driven oversight is emerging as a best practice for responsibly scaling AI in medicine. Read full story.
Harvard’s BKC Explores Whether Human Intelligence And AI Computational Intelligence Are Actually The Same
Author: Lance Eliot Published: September 28, 2025
Harvard’s Berkman Klein Center hosted discussions on AI versus human intelligence. The consensus was that human thought is not literally the same as AI computation, but effective AI must emulate certain aspects of human problem-solving. The center explored how understanding human cognition can inform better AI design and vice versa, but ultimately concluded they remain distinct phenomena. Read full story.
