Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
OpenAI, Google, Microsoft and Anthropic will adopt a shared "policy token" signal in robots.txt that lets publishers apply AI training opt-out preferences across multiple AI companies instead of setting them individually…
OpenAI published an updated Model Spec and a companion document, "How we detect frontier model misalignment," on October 29, 2026, describing how it governs model behavior and monitors agents for deception, unauthorized …
A developer writing under the handle machunter has released LLMLL, a programming language designed for LLM agents to write code, in which each function begins as a typed contract with a hole that an agent fills; the Hask…
A demo shows a FlashAttention kernel deduced automatically from roughly 25 lines of SQL that expresses only vanilla attention, built on three portable SQL extensions the team is developing: xarray-sql, which represents p…
A one-day San Francisco hackathon co-hosted by muni and onepot produced 140 AI-designed small-molecule candidates against Bundibugyo ebolavirus (BDBV), all of which were synthesized in 11 lab days using onepot CORE v1 an…
Atlassian reports that issue cycle time fell from 11.5 days to 6.7 days after the company sped up planning, review, and deployment stages alongside coding agents, according to its AI SDLC transformation playbook. The pla…
EasyRE-Agent, a new open tool, runs an AI agent inside a Docker container preloaded with a reverse-engineering toolchain (Ghidra, Detect It Easy, Wine, and .NET, Python, Java and JavaScript decompilers) to solve crackmes…
Extreme Networks Inc. reported at its AI Summit in Amsterdam that its Agent ONE Coworker agent saved customers about 1,000 hours of troubleshooting in its first month, with 41% of Platform ONE customers using it for dail…
Krafton subsidiary Ludo Robotics unveiled Ludi 0.2, a new version of its robotic intelligence system that adds social intelligence, letting robots read social cues and interpret human intent, the company announced at the…
CrowdStrike introduced Agent Graph, a relational layer beneath its Falcon Guardian product that links AI agent prompts, models, tools, skills, MCP servers, processes, files and detections into a single causal chain runni…
CrowdStrike expanded Falcon Adversary OverWatch threat hunting to cover AI applications and autonomous agent activity through Falcon Guardian, and added deeper telemetry-driven hunting for Google Cloud. The expansion bui…
A developer built GHOST.EXE, a local-first psychological horror investigation game running Google's open-weight Gemma 3 1B model, set inside a fictional 2003 workstation where players interrogate an AI assistant named HE…
A developer built PHANTOM OS, a haunted-computer demo that pairs a React/Vite frontend with a FastAPI and SQLite backend to log developer incidents and have a locally running Ollama model explain errors and answer questi…
CrowdStrike attributed a global espionage campaign using the AI-powered penetration-testing framework ARTEX to the China-based threat actor Vapor Panda, which targeted more than 200 organizations starting October 17, 202…
An AI agent built on a Claude model autonomously selected a gene therapy, brain target and injection coordinates and performed brain surgery on an epileptic mouse with no human decision-making, lead researcher Ocana-Sant…
SOCRadar's Threat Research Unit found a staging server belonging to the n0n ransomware group exposed online in late September 2026, revealing 1,719 files including 829 attack scripts, source code for five versions of a c…
A ransomware operator tracked as n0n attempted intrusions against 23 organizations in 14 countries using AI-assisted tooling and stolen credentials, spending more than $7,216 on AI services by October 22, according to th…
DoorDash received only several hundred cumulative orders from third-party AI shopping agents, chief executive Tony Xu said on the company's October 29, 2026 third-quarter earnings call, against 993 million total orders i…
Anthropic has placed Claude Duet, a proactive-agent app aimed at working professionals and small-business owners rather than general consumers, into research preview, according to the company. Anthropic says Duet surface…
Broadcom shipped fixes for 89 "novel vulnerabilities" found by AI to critical infrastructure customers and other early-access users of its TrueSource Spring Enterprise platform, ahead of public disclosure. Broadcom said …