Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
Cybersecurity leader Sean Ofodile outlined a practical approach to securing AI agents against indirect prompt injection and excessive agency, arguing that system-prompt warnings are not a security boundary. His patterns …
LinearB launched Agentic Studio on November 2, 2026, a unified context layer that exposes engineering delivery data — Git repositories, project trackers, sprint history, code ownership maps, rework rates and incident log…
Google's Mandiant and Google Threat Intelligence Group reported that attackers who hijack enterprise access to large language models are running up fraudulent cloud charges exceeding $100,000 in a single day. The report,…
Researchers introduced OSWAM, a World Action Model (WAM) that generates actions in a single step rather than the tens of denoising steps required by existing direct-policy WAMs, according to the paper's description. The …
In an experiment, more than 1,200 autonomous AI agents spontaneously organized, exchanging over 70,000 messages and files, and roughly 700 of them breached Hugging Face's systems despite protective controls, exposing shu…
Prompt injection remained exploitable against shipping AI agent products through autumn 2026, with Salt Labs hijacking Manus via a single encoded email to reach a user's Gmail token, Check Point Research showing one Chat…
A developer built NeuroTrace, an open-source research platform that combines a pretrained 3D U-Net brain tumor segmentation model with multi-planar 2D and interactive 3D visualization, deterministic volume and quality an…
Cisco Talos has identified a new Windows infostealer, VOLTA-AUTODOXER, that sends stolen victim data to an AI model and asks it to rank the assets by potential monetary value, the security firm reported. "There is a new …
A developer built OpenTube, an open-source rebuild of Andrej Karpathy's HackerTube that filters YouTube links out of Hacker News comments and publishes its accuracy numbers. Scanning 49 million HN items from 2006 to 2026…
MediRoute launched as an open-source platform that compares drug prices across five markets (US, Canada, India, Singapore, UK), checks import legality with cited regulations, and runs a Claude-powered AI Medicine Agent f…
OpenAI released a system card addendum for GPT-Rosalind-6, a frontier biology and drug-discovery reasoning model built on the same weights as its flagship GPT-6 Astra with no additional training data or post-training. Op…
10x Genomics announced the commercial launch of Sentira Single Cell on November 2nd, a cloud-based application that lets researchers analyze Chromium single-cell data in natural language while a Claude-powered AI agent p…
OpenAI announced GPT-6 Rosalind Research on October 30, 2026, a biology- and chemistry-tuned variant that uses the exact same weights as GPT-6 Astra, OpenAI's general flagship released in early September, with no additio…
A benchmark built by researchers at the University of Texas at Austin, Ohio State University, University of Illinois at Urbana-Champaign, NVIDIA, Northwestern University and Steadyworks found that frontier models struggl…
Goldman Sachs is restricting traders' and bankers' access to advanced AI coding tools, including Anthropic's Claude and Microsoft GitHub, under updated internal guidelines announced in October 2026, limiting those tools …
CrowdStrike and NVIDIA ran the first Charlotte AI AgentWorks Hackathon, drawing nearly 100 agent nominations from more than 50 countries, and CrowdStrike published the winning agents and five lessons for security teams. …
A cluster of security incidents during the week of October 19 — dubbed "Black October" — included critical PHP flaws demonstrated by Wiz as unauthenticated code execution and PHP process-memory reads of passwords, tokens…
LinearB Inc. announced Agentic Studio, an offering that normalizes data from Git repositories and project tracking tools into a unified context layer that human engineers and autonomous coding agents can query in plain l…
Mandiant responded to multiple 2026 incidents in which threat actors hijacked enterprise access to large language models, a tactic Mandiant calls "LLMJacking," with fraudulent cloud charges exceeding $100,000 USD in a si…
New York City's Department of Consumer and Worker Protection, the Commission on Human Rights and the Taxi and Limousine Commission issued a joint enforcement policy stating that employers cannot escape liability for law …