What happens in AI chat doesn't stay in AI chat. Oops, all your secrets are no longer secrets, a researcher found: - Evidence that at least three AI providers were compromised by attackers who used their access to siphon prompts they received - Multiple resellers in the inference marketplace that logged the prompts and somehow passed them along to people who used the credentials - Multiple cases of AI providers not following end-user privacy settings embedded into flags specifying terms such as Zero Data Retention (ZDR) or “no data collection” arstechnica.com/security/2026/… #LLM #Privacy #Security #AI
AI · News · page 2
Further down today’s ranking of AI posts that are news, facts and calls to action: page 2 of 2. Posts the model is less sure belong here come last.
“These systems are solving problems in a few hours that I thought I’d be working on for years,” said a postdoctoral researcher in statistics. “It’s hard on the ego," Winer said. "But I’m much more concerned about the Mike-Winer-and-everyone-I-know-dying timeline.” www.nytimes.com/2026/11/22/m...
Claude Code’s persistent permissions have exposed credentials through configuration files. A reminder that AI agent security extends beyond code: permissions, execution history, and configuration are part of the attack surface. www.infoq.com/news/2026/11... #AI #Security
A new preprint (GAMA-Bench, 1,298 matched scenarios) found 10 leading AI models judged men more harshly than women in identical conflicts: more blame, less empathy, even with step-by-step reasoning. Tip: using AI to review complaints or HR issues? Swap the genders and compare. #AI #AIbias
Closing the loop: Ashby's ultrastable system and why today's AI systems disappoint A common belief: AI systems are smart enough already; what holds them back is model capability. What holds most AI systems back is an open loop: nothing checks the result against l [AI-generated]
A couple of teams made a point of implementing #LocalAI and one was quite excited to get #Apertus working in their financial advisor app 🫕 #powercoders #hackathon > We built TellHorn, a platform that integrates with your PostFinance app to bring real clarity to your spending, pairing your bank feed with receipt-level detail, and uses AI agents to actively optimize your expenses, spot hidden leaks, and find concrete ways to save money. github.com/TenzinJhopee/HackPo…
AI accelerates scientific research? NVIDIA just announced Rubin HPC, a new GPU aimed specifically at scientific computing, engineering, analytics, and AI-driven simulations. What’s particularly interesting is the focus on AI agents + simulation: agents can generate ideas, run simulations, evaluate results, and iterate faster. Rubin HPC is expected to be available in 2028. blogs.nvidia.com/blog/rubin-hp… #NVIDIA #RubinHPC #AI #HPC #GPU #AIInfrastructure #AcceleratedComputing #DataCenter #technology
AI clusters are limited less by chip design than by the release pipelines pushing firmware and drivers across tens of thousands of accelerators. An infrastructure lead describes a six-day firmware rollout frozen by a false alarm, Meta's Llama 3 runs logging 466 job interruptions in 54 days on 16,000 H100s, and fleet-wide releases taking two weeks. venturebeat.com/ai/even-the-fa… #Tech #Startup #News #AI
Keynote Dr. Jodie Burchell talking about AI is having its moment ... again #PyConSweden2025 #PyConSweden
"the phenomenon of experts reading newspaper articles within their fields of expertise and finding them to be error-ridden and full of misunderstanding, but seemingly forgetting those experiences when reading articles in the same publications written on topics outside of their fields of expertise, which they believe to be credible." en.wikipedia.org/wiki/Michael_… This post is about AI
"AI is great at writing code. But it doesn’t know if you are launching the app or have millions of users, which it needs to make the right architecture call to log a timestamp vs email. It is also sycophantic and tries to execute your requirements exactly without pushing back, which means you end up with a slow app that takes 15 seconds to load a static welcome screen." manidoraisamy.com/developer-to…
Books written with generative AI are hitting the market in such abundance that studies show they are hurting the revenue of human authors armed only with the power of imagination. blog.ecomtechnology.com/2026/1…
Daniel Miessler has created AI Hacking Incidents, a public ledger of all known AI-caused security incidents hackingincidents.ai/
When someone forms a close relationship with an AI companion, what happens when it ends? New interviews reveal how people experience grief, attachment and loss in digital relationships. www.theguardian.com/technology/n... arxiv.org/abs/2609.14696 #DigitalSociology #AIRelationships
[OT] Turing completeness of LLMs? (2024) medium.com/heyjobs-tech/turing… "One of the clearest differences between LLMs and a Turing machine is that LLMs are probabilistic, not deterministic." #AI #LLM
Researcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced. arstechnica.com/security/2026/…
"I docenti con l’intelligenza artificiale risparmiano tempo, ma devono controllarla con competenza: intervista al prof Luciano Floridi" tecnicadellascuola.it/i-docent…
Experts warn: Nearly everyone in AI is susceptible to a honey-trap scheme right now techcrunch.com/2026/11/20/expe…
More in AI
The model is less sure these belong here, so they sit outside the ranking.
Anthropic researcher says he and Claude proved the Riemann hypothesis, first posed in 1859 Source : Phys. org / Alex Pigman with Thomas Urbain phys.org/news/2026-11-anthropi… #mathematics #maths #math #AI
KoboldCPP 1.125.1 has been released by creator Lost Ruins as an open-source tool for running AI models locally and entirely offline without cloud connections or data sharing. The release supports CPU, NVIDIA CUDA, and Intel oneAPI acceleration, running on Windows, Linux, and macOS systems including Apple Silicon M-chips. It requires at least 8GB of RAM with 16GB+ recommended, and works exclusively with.gguf model files while ser… en.hacks.gr/koboldcpp-1-125-1/ #KoboldCPP #LocalAI #OpenSource #Linux