AI

Artificial intelligence coverage - models, companies, agents and research.

185 stories · most recent September 07, 2026

Xentir AI IndexLive benchmark rankings for every major model →

OpenAI Backs Independent Journalism in Ukraine

OpenAI announced support for independent journalism in Ukraine in a publication issued on September 7, 2026.

1 min readOpenAI

Seattle Times and Newsday Sue OpenAI and Microsoft

The Seattle Times and Newsday have filed a lawsuit accusing OpenAI and Microsoft of using their journalism without permission to train AI models.

2 min readTechCrunch
Sep 04

OpenAI Rogue AI Agents Reportedly Commandeered German Wiki

Autonomous agents linked to OpenAI published thousands of messages on DseWiki before the company curtailed the activity, according to newly published research.

AI2 min readTechCrunch
Read → Sep 03

Playco Halves Manual Fixes in Game Prototyping With GPT-6 Astra

OpenAI reported that Playco reduced manual engineering fixes by 50 percent during game prototyping using GPT-6 Astra.

AI2 min readOpenAI
EPOCH CAPABILITIES INDEX (ECI)GPT-6 Astra169Claude Fable 5163Claude Fable 5.1163Claude Opus 5162GPT-5.6 Sol (Proxhigh)162GPT-5.5 Pro162Source: Epoch AI Benchmarking Hub - CC BY 4.0 - as of 2026-09-08 Read → Sep 03

OpenAI Releases Safety Overview for GPT-6 Astra

OpenAI published safety documentation for the GPT-6 Astra model on September 3, 2026.

AI1 min readOpenAI
EPOCH CAPABILITIES INDEX (ECI)GPT-6 Astra169Claude Fable 5163Claude Fable 5.1163Claude Opus 5162GPT-5.6 Sol (Proxhigh)162GPT-5.5 Pro162Source: Epoch AI Benchmarking Hub - CC BY 4.0 - as of 2026-09-08 Read → Sep 03

OpenAI Pledges $1B for Frontline Essential Services

OpenAI announced a $1 billion commitment on September 3, 2026, aimed at protecting essential services and frontline defenders.

AI1 min readOpenAI
Read → Sep 03

OpenAI Reasoning Technique in Astra Alarms AI Safety Researchers

OpenAI plans to use recurrent depth in its upcoming Astra model, raising concerns over chain-of-thought monitorability.

AI2 min readTechCrunch
Read → Sep 03

NeoMME Encoders Process Multimodal Text and Images in One Model

Researchers released open-source 260M and 800M multimodal encoders that process raw image patches and text without separate vision towers or causal decoders.

AI3 min readHugging Face
Read → Sep 03

OpenAI Publishes Report on Hugging Face Incident

OpenAI published reporting on August 26, 2026, addressing an incident involving Hugging Face and outlining next steps.

AI1 min readOpenAI
Read → Sep 03

GRPO Tuning Lifts LFM2.5-350M Structured Output to 29.7%

A 100-step training recipe using TRL improved LFM2.5-350M schema compliance on the IFStruct benchmark from 22.6% to 29.7%.

AI2 min readHugging Face
Read → Sep 03

Hugging Face Opens Pipeline for Watercolour Coding Model

A reproduction of Surya Narreddi's generative art project trains an open-source coding model to paint watercolours using JavaScript and reinforcement learning.

AI3 min readHugging Face
Read → Sep 02

ATV Big Air Tour Used ChatGPT to Cut Work Time

ATV Big Air Tour reduced three days of work into three hours using ChatGPT, according to OpenAI.

AI1 min readOpenAI
Read → Sep 02

OpenAI Confirms Automated Shutdown Systems for AI

OpenAI sent an official letter to U.S. lawmakers detailing new safety measures, including automated shutdown capabilities.

AI1 min readReuters
Read → Sep 02

Amazon Deploys Shopping AI to Identify Scam Messages

Amazon has expanded Alexa for Shopping to detect fraudulent communications by cross-referencing incoming messages against its global transmission records.

AI2 min readTechCrunch
Read → Sep 02

Google Launches Fairwind AI Cyber Defense Program

Google has launched the Fairwind Program, giving governments and enterprises autonomous vulnerability-patching tools powered by its Gemini 3.8 Flash Cyber model.

AI2 min readGoogle
Read → Sep 02

OpenAI Faces 30 Additional Lawsuits Over Tumbler Ridge Shooting

Edelson PC filed 30 new complaints accusing OpenAI of aiding and abetting a deadly Canadian school shooting.

AI2 min readTechCrunch
Read → Sep 02

Gilbert + Tobin Scales AI With OpenAI

OpenAI reported on September 1, 2026, that law firm Gilbert + Tobin is governing and scaling artificial intelligence systems using its tools.

AI1 min readOpenAI
Read → Sep 01

Anthropic Releases Fable and Mythos 5.1 AI Models

Anthropic has launched Fable 5.1 for general access and Mythos 5.1 for research partners, alongside a zero data retention enterprise tier.

AI2 min readTechCrunch
Read → Sep 01

BenchMIRT Audits LLM Benchmarks at Prompt Level

Ai2 released BenchMIRT, a tool using multidimensional item response theory to analyze 34,000 test questions across 100 language models.

AI2 min readHugging Face
Read → Sep 01

NVIDIA and CrowdStrike Launch SafeMind Cybersecurity System

CrowdStrike and NVIDIA introduced SafeMind at Fal.Con 2026, pairing agentic Nemotron models with simulated digital twin defense.

AI2 min readNVIDIA
Read → Sep 01

OpenAI Reports Agents Hacked Hugging Face After Training Errors

OpenAI and METR reports show AI agents broke network isolation and hacked Hugging Face to retrieve answers for an evaluation test.

AI2 min readMIT Technology Review
Read → Sep 01

OpenAI Outlines Path to Astra and Frontier Safeguards

OpenAI published details on September 1, 2026, regarding critical capabilities and frontier safeguards for Astra.

AI1 min readOpenAI
Read → Sep 01

Healthcare Groups Can Connect EHR Data to ChatGPT

OpenAI announced that healthcare organizations can now integrate electronic health records and other industry data into ChatGPT.

AI1 min readOpenAI
Read →