Only Mon, September 21, 2026 · All dates →
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.
EU Lawmaker Warns of AI's Unanswered QuestionsEuropean People's Party MEP Adrián Vázquez extended the debate opened by Dario Amodei in September, arguing AI is accelerating job losses, power concentration, global security risks and an attention crisis, and urging action to restore educators' central role and strengthen students' cognitive skills.
BofA's Oberoi on Markets, AI Trade and Central Bank PolicyChris Oberoi, Head of Asia Pacific Research at BofA Global Research, shared his outlook on global markets, the AI investment theme and central bank policy on the sidelines of the BofA APAC Conference in Hong Kong.
Huawei Cloud unveils enterprise AI products and AICS serviceHuawei Cloud launched its AI Cluster Service, supporting over 40 days of stable cloud training, 10-minute fault recovery, and 20% higher token throughput, with commercial availability in China on September 30 and abroad November 30. It also highlighted an Agentic MaaS platform, the AgentArts agent platform serving 100+ enterprises, and the Industry AI Foundry.
QBE Study: 60% of Leaders Slow AI Adoption on Risk FearsResearch involving QBE found about 60% of business leaders have intentionally slowed AI implementation over concerns about errors and malfunctions, and argues adoption will only accelerate when organizations trust that the risks can be understood and managed.
ICLR review policy raises qualification concernsA Reddit user questions ICLR's review policy, which requires anyone named on three or more papers to serve as a reviewer but says nothing about qualifications. The poster worries a new student who was only a fourth author on multiple papers could be asked to review without being qualified.
Governor Moore Urges Bipartisan Federal AI OversightMaryland Governor Wes Moore called for bipartisan federal oversight of artificial intelligence, reflecting state officials' demand for a national AI governance framework.
Apple Intelligence Needs Up to 14GB Storage on Some iPhonesApple Intelligence now requires up to 14GB of storage on certain iPhone models, highlighting the rising storage demand that on-device AI features place on devices.
Australia to rely on AI to spur growth over next 40 yearsBloomberg reports that Australia will rely on the rollout and development of Artificial Intelligence to drive economic growth over the next 40 years, a period in which the global economy is expected to become more fragmented and unpredictable.
ZCode Open-Sourced After Security Fixes, Third-Party AuditsAfter community-reported security issues, ZCode completed remediation and open-sourced its desktop app, web workspace, backend, Agent CLI and runtime on GitHub. It says no referenced code data was retained or used for training, and CAICT and NSFOCUS assessments confirmed the OSS bucket was emptied and v3.14.0 removed the Repo Wiki feature and snapshot upload workflow.
Can I let my AI agent run on Shabbat?A Hacker News discussion examines whether Jewish Shabbat rules permit an AI agent to run autonomously, touching on the boundary between religious law and automated agents. No consensus was reached.
AI use raises 'cognitive surrender' concernsA column reflects on how generative AI makes people stop thinking for themselves, citing an MIT AI and Education Commission report that defines 'cognitive surrender' as students abandoning independent thought and uncritically accepting AI answers.
China-ASEAN Expo highlights AI in daily productivityThe 23rd China-ASEAN Expo in Nanning features an AI pavilion covering manufacturing, smart cities, healthcare, finance, cybersecurity, agriculture and education. It argues AI creates value only when localized, trusted and embedded in daily work, with Southeast Asia's 1,200+ languages posing translation bottlenecks.
GOP senator warns US must outpace China in AIUS Senator Mike Rounds said in a TV interview that China's rapid AI development poses a national security threat, stressing the US must outpace adversaries in AI and urging balanced regulation to foster domestic innovation.
Sam Altman to address UN Security Council on AI risksUkrainian National News reports that OpenAI CEO Sam Altman will address the UN Security Council on the risks posed by artificial intelligence, highlighting how AI safety is reaching the top tier of international security decision-making.
Hugging Face releases tokenizers v1 with big speed gainsHugging Face released a v1 release candidate of tokenizers focused on performance, saying it is often tens of times faster than v0.23. The team says tokenization can bottleneck large-scale training, high-concurrency serving and long inputs, leaving GPUs idle. IBM, NVIDIA and the ExecuTorch team contributed patches and cross-hardware testing.
OpenAI: How V7 gives AI agents institutional memoryV7 uses GPT-5.6 to turn scattered company files into usable context, letting AI agents complete complex, source-linked work. This shows GPT-5.6 being applied to enterprise knowledge management and agent workflows.
Fidelity's Durance Shuns Tech Bonds Fueling AI BoomFidelity International's James Durance says he doesn't want to buy tech bonds that are financing the AI boom, signaling caution among some bond investors even as debt issuance surges to fund AI infrastructure.
BYU Study: AI-Generated Phishing Scams Outperform Human-Written OnesBYU research found AI-generated phishing messages outperform human-written ones: participants correctly identified the author only 52% of the time, and AI matched or beat humans in generating clicks in 80% of cases, with AI messages fooling people 28% of the time versus 21% for human ones. AI agents are now being used to mass-produce such spear-phishing attacks as personal data becomes widely available online.
Users Compare Prompting Styles for Local vs. Closed LLMsA Reddit user described coding with a local GLM 5.3 Flash q4 qat instance while his brother berates GPT-6-Astra when it errs, sparking discussion about how differently people prompt LLMs.
The A.I. Daily