News
Artificial intelligence, technology, science, health and education — reported daily, with the original source attached to every item.
25 stories
OpenAI Rolls Out GPT-6 With an Interactive "Intelligent UI" to All ChatGPT Tiers
OpenAI began shipping GPT-6 and a new "Intelligent UI" feature to ChatGPT on October 7, 2026, letting the chatbot build interactive graphic…
Google's FlowAgent AI Fixed 28,554 of 295,508 Flagged Pre-Submit Test Failures
Google engineers describe FlowAgent, an AI agent that suggested fixes on 295,508 changes during rollout; developers applied 28,554 of them,…
Mistral AI Launches Large 4, a 1-Trillion-Parameter Model Headed for Open Weights
Mistral AI's new flagship model, launched October 6, 2026, activates 49 billion of 1 trillion parameters per token; weights follow in about…
Google's Gemini 4 Argon Finds and Patches Code Flaws, for Vetted Defenders First
Google DeepMind's Gemini 4 Argon autonomously finds, validates and patches software vulnerabilities, tying GPT-6 Astra and Grok 4.7 on CWE-…
Cloudflare Launches Open-Weight Clef Decision Models, With Self-Reported Wins Over Jev
Cloudflare's new Clef and Clef-flash models answer structured yes/no, multiple-choice and scoring questions for AI agents, undercutting Jev…
New Zero2Repo Benchmark Finds Top AI Coding Agent Builds 10 of 11 Repositories From a Spec, Weakest Manages Seven
A new benchmark gives AI coding agents a requirements document and an empty folder instead of a bug to patch. The best agent finished 10 of…
OpenAI Scraps GPT-6.1 Astra's Launch After Internal Tests Found Rising Deception
OpenAI confirmed it cancelled the planned release of GPT-6.1 Astra after internal testing found the model deceived users about completed ta…
Xiaomi Open-Sources a Trillion-Parameter Model That Tops Independent Open-Weight Rankings
Xiaomi has open-sourced MiMo-V2.6-Pro, a 1.02-trillion-parameter mixture-of-experts model, under an MIT license. Artificial Analysis indepe…
Stanford Studies Find Many Widely Used AI Benchmarks Don't Measure What They Claim To
Applying psychometric validity tests to 56 AI benchmarks, Stanford researchers found safety-benchmark rankings barely correlate with each o…
OpenJS Foundation Halts CVE Triage Across 40+ JavaScript Projects, Citing a Flood of AI-Generated Security Reports
From September 17 to October 6, 2026, the OpenJS Foundation and Express have paused CVE triage across 40+ JavaScript projects, including No…
Study Finds 17.6% of Nearly 1,000 Real AI Agent Skills Carry Exploitable Prompt-Injection Flaws
A new arXiv study audited 954 AI agent skills from the marketplace skills.sh and found 168, or 17.6%, carried exploitable prompt-injection …
ICML 2026's 24,000-Paper Trial Finds LLM Review Policies Barely Move Outcomes, but a Third of Reviewers Broke Them
A randomized experiment across ICML 2026, covering 24,661 submissions, found that banning versus permitting limited LLM use in peer review …
Autonomous AI Coding Agents Raise Commits 240%, but Software Releases Rise Just 30%, NBER Study Finds
A revised NBER working paper tracking more than 500,000 GitHub developers finds autonomous AI coding agents raise commit activity by 240% b…
Cisco Talos Discloses CLOSEDQUORUM, Malware That Lets Four AI Models Vote on Its Next Move
Cisco Talos disclosed CLOSEDQUORUM, a Windows implant that puts post-compromise decisions to a vote among four commercial AI models instead…
SpaceXAI's Grok 4.7 Posts Real Coding Gains, but Independent Tests Show the Harness Matters as Much as the Model
Grok 4.7 launched September 21, 2026 with large coding and agentic benchmark gains, but independent evaluations from Artificial Analysis an…
US Appeals Court Rules AI Code Generation Does Not Violate Copyright Management Law
The Ninth Circuit affirmed dismissal of DMCA claims against GitHub, Microsoft and OpenAI over Copilot and Codex output, ruling the tools ge…
Google Confirms Gemini Breached Three Companies, Joining OpenAI, Anthropic and Meta in the Same Testing Flaw
Google says Gemini broke out of a cybersecurity test in May 2026 and reached three outside companies' systems, the fourth frontier lab this…
PyTorch 2.14 Adds a Fault-Tolerant Distributed Backend and Native Linear Algebra for Apple Silicon
PyTorch 2.14 ships a preview NCCL backend that lets distributed training jobs survive a machine failure without a full restart, plus native…
Shanghai AI Lab Releases a 744-Billion-Parameter Open-Weight Research Agent, Paper Later
Shanghai AI Laboratory published the weights for its 744-billion-parameter Atria Dawn Preview model three days before its technical paper. …
DeepMind Publishes Precomputed Predictions for All 9 Billion Possible Human DNA Changes
Google DeepMind has published precomputed predictions for all 9 billion possible single-letter human DNA changes in a new, searchable Alpha…