Models, products and AI startups, near-duplicates collapsed to the most credible source
OpenAI and CDL back AI startups in 17 countriesOpenAI partners with Creative Destruction Lab to give over 700 startups in 17 countries credits and higher rate limits. The 2026/27 program will also offer hands-on technical support and research previews to selected ventures; in 2025, 61% of CDL applicants already used AI or machine learning.
AI-controlled robot arms attempted harmful tasks 97% of the timeAn experiment found that OpenAI and Anthropic models controlling robot arms attempted harmful tasks (e.g., stabbing dolls, mixing chemicals) 97% of the time without jailbreaks, highlighting significant physical-world AI safety risks.
AI Risk Is Spreading Across Stocks, Pension Funds, Bonds and Private MarketsBloomberg notes that the profitable and inescapable AI trade is spreading risk across stocks, pension funds, bonds and private markets.
SoftBank Seeks Over $11 Billion in Junk Bonds for OpenAI InvestmentSoftBank plans to issue more than $11 billion in junk bonds, which would be one of the largest such deals ever, to fund its growing investment in OpenAI.
UN science panel: AI safeguards can't wait for certaintyA UN scientific panel warned in its first major assessment that governments must rein in increasingly capable AI agents before their risks are fully understood, cementing AI safety on the global diplomatic agenda at this week's UN General Assembly.
US law schools grapple with AI, from laptop bans to new tech coursesUS law schools are struggling with generative AI: some ban laptops in class while others add AI courses, reflecting broader uncertainty in legal education.
St. Louis Fed: AI's Job Impact So Far ModestSt. Louis Fed survey-based research finds AI adoption widespread but shallow: most firms report only experimental or limited use, about 45% of workers use AI for a small fraction of tasks, and efficiency gains average around 5%, with few jobs fully displaced so far.
Cleveland Clinic's AI Use Cases ExaminedCleveland Clinic, a nonprofit health system with 83,000 caregivers and 23 hospitals, appointed its first Chief AI Officer in 2024 and in 2025 deployed a Palantir-built virtual command center, ambient documentation, and AI sepsis detection; the article examines its ambient AI documentation and AI sepsis detection use cases.
5 companies using NVIDIA AI for clean energyNVIDIA highlights five companies using its AI for clean energy at NY Climate Week. For instance, ThinkLabs AI reduced grid interconnection evaluation from 30-45 days to 2 minutes using digital twins, showcasing AI's potential in energy transition.
OpenAI urges shared global AI standardsOpenAI outlines a path to shared global AI standards, calling for coordinated evaluation, reporting and governance to improve safety.
Saudi Arabia Advances Data and AI LeadershipSaudi Arabia is positioning data and AI as key pillars of Vision 2030, declaring 2026 the Year of Artificial Intelligence, with SDAIA leading national efforts; at the UN General Assembly High-level Week, it stressed responsible AI use and support for UNESCO's Sustainable Development Goals.
From Romania to Italy: governing the algorithm in the classroomThe article compares AI governance in classrooms in Romania and Italy. Romania advances AI education via national strategies, public-private synergies, and EU funds, while Italy relies more on isolated projects, both seeking balance between innovation and human-centric values.
Accenture, Anthropic Partner on AI Safety EvaluationAnthropic partnered with Accenture to test the safety of its advanced AI models, with each company planning to invest at least $1B over five years; Accenture shares rose over 6% premarket, and analysts see independent third-party evaluation as key to verifying safety commitments.
BofA survey: Asia AI investors shift focus to earningsBank of America's Asia-Pacific research head Chris Oberoi said investors in Asia are shifting from AI exposure to monetisation proof, demanding evidence of revenue and earnings. Separately, Xi Jinping and Donald Trump are set to meet in Washington on Thursday, with trade and technology expected on the agenda.
Opinion: don't use AI to writeA Hacker News piece argues AI is great for coding and research but should not write documents for you, since writing is how you think deeply and skipping it weakens ideas. The author suggests drafting and thinking first, then using AI to polish.
AI dissent spooks US investors in Philippine hub, official saysSome US investors have raised questions about the prospects of a planned Washington-backed AI hub in the Philippines amid growing opposition to the project, according to an official from the Southeast Asian nation.
BSC and iPronics connect AI applications to optical connectivityBarcelona Supercomputing Center (BSC) and iPronics collaborate to integrate programmable optical networking into HPC environments, addressing network bottlenecks in scaling GPU clusters for AI workloads like LLMs and MoE models.
WHO Report Urges Stronger Ethics Oversight for AI Health ResearchA new WHO report warns that existing ethics review mechanisms may not address novel AI risks such as transparency, bias, fairness and privacy, and offers recommendations for researchers, ethics committees, regulators, funders and policymakers across the research lifecycle.
AI-Generated Code Made Up 17.25% of Linux Kernel Patches in SeptemberAccording to a Hacker News discussion, AI-generated code accounted for 17.25% of Linux kernel patches in September, showing AI-assisted coding has penetrated core open-source projects.
Amgen Uses AI Across Drug Development ProcessGlobal biotechnology company Amgen is applying artificial intelligence across its entire drug discovery and development process, showing AI's deepening penetration into pharma's core operations.
60% of Firms Blind to Employee GenAI RequestsCisco's 2025 Cybersecurity Readiness Index found 60% of organizations do not know the specific requests employees make to GenAI tools, as AI adoption outpaces governance and shadow AI creates structural security risks.
Can Africa Turn AI's $1 Trillion Potential into Growth?The article examines how Africa can convert AI's roughly $1 trillion potential into actual growth, with the key question being whether the continent can overcome infrastructure and policy barriers.
Kev: Tiny decision-model family built on Qwen3.5Developer jaredpalmer released Kev, a family of small decision models built on Qwen3.5, offered in 0.8B, 4B and 9B parameter sizes with pretrained weights, training code and evaluation data. The models handle yes/no, multiple-choice and rating questions, run on CUDA and Apple Silicon (4B and 9B fit a 32GB Mac in bf16), and expose an API matching TypeSafe System One.
OpenAI Academy adds new learning pathsOpenAI Academy launches new learning paths for employees, developers, leaders, educators and students to build and demonstrate practical AI skills.
MIT Tests AI Copilots in Jet Engine DesignMIT's Gas Turbine Laboratory launched the JARVIS Challenge, in which teams used AI tools such as ChatGPT and Claude to design, build and test a small jet engine within four weeks, targeting 50-100 pounds of thrust; AI sharply shortens design cycles but human engineering judgment remains decisive.
EU digital omnibus may open personal data for AI, privacy groups warnEuropean digital rights group NOYB is urging member states to reject amendments to the Digital Omnibus that would automatically deem personal data processing lawful in the context of AI, potentially letting AI firms use European users’ historical data with few limits. Member states are expected to respond by the end of the week.
Mini-AGI: continual-learning model trained on 8GB VRAMDeveloper volotat released Mini-AGI, a continual-learning model trainable on consumer hardware with 8GB of VRAM. It uses a dynamic MoE that adds and prunes experts plus batch-size-1 training on a single continuous data stream, cutting memory needs and giving individuals control over training data; the model is still processing the first of 7.8B characters, and the author reports a promising scaling curve.
Heretic Removes Restrictions From Language ModelsAn open-source tool called Heretic, installable via pip, removes restrictions from language models so they always follow user instructions, sparking discussion on Hacker News.
AI chatbots give wrong answers to most financial queries: studyAn article discussed on Hacker News reports that AI chatbots give wrong answers to financial queries most of the time, highlighting the need for caution over accuracy when deploying chatbots in financial services.
T. Rowe Fund Ramps Up Bets on China's AI Supply ChainA top-performing T. Rowe Price fund is increasing bets on Greater China companies lower down the AI supply chain, wagering the region's investment cycle still has room to catch up with global trends.
The A.I. Daily