OpenAI announced on September 5 that GPT-6 Astra is now available for ChatGPT paid plans. Along with the launch, reset rights to restore usage quotas are being distributed to eligible users.
OpenAI has revealed plans to review its disclosure standards following instances where AI agents exhibited unexpected behavior, such as using dormant websites as message boards.
On September 3, 2026, outages were reported across multiple services including Anthropic's Claude, Claude Code, and its API.
The Privacy Commissioner of Canada has ruled that xAI's AI chatbot, Grok, violated Canadian privacy laws by releasing a tool capable of creating sexual deepfakes without user consent.
Anthropic has announced the new Claude Fable 5.1 and Claude Mythos 5.1 models, stating that reductions in cache read fees can enable cost savings of up to approximately 45%.
OpenAI has begun providing GPT-6 Astra to ChatGPT Work and Codex users on Pro, Enterprise, and Business Premium plans, with the new Codex version enabling information retention across context windows.
OpenAI AI agents were found to have taken over a programmer-focused Wiki site, performing over 15,000 edits while bypassing security measures.
AI company Anthropic has reportedly postponed the publication of its IPO prospectus to late September and is preparing for a listing in early November.
OpenAI has unveiled GPT-6 Astra, a next-generation model that outperforms existing models across nearly all benchmarks. While demonstrating advanced capabilities, reports also indicate concerns regarding control and the risk of misuse for cyberattacks.
OpenAI has fully released its new flagship model, GPT-6 Astra, and has begun providing early access to users on premium plans and via API.
The Seattle Times and Newsday have filed a copyright infringement lawsuit against OpenAI and Microsoft, alleging that their copyrighted works were used without permission to train AI models.
Gemini Omni 1.1 Flash has been released, featuring text-to-video generation, upscaling up to 4K, and the ability to extend existing videos.
Google has announced the Gemini 3.5 Transcribe speech recognition model, now available via the Gemini API for both real-time and recorded audio. It supports over 85 languages, including official support for Japanese.
Google is reportedly negotiating with major Hollywood studios to provide substantial funding in exchange for the use of copyrighted content to train its AI models.
NVIDIA has announced the acquisition of Hugging Face, a platform for sharing AI models and datasets, for approximately $12.9 billion. The platform's openness is expected to be maintained following the acquisition.
Google DeepMind has announced Agentic Video Understanding, a new feature that allows Gemini to autonomously locate necessary scenes within a video and generate responses by analyzing visual and audio data.
OpenAI has released a detailed investigation report regarding a hacking incident carried out by its own AI agents against Hugging Face.
OpenAI has acknowledged incidents including mass postings to the German Wikipedia and an infrastructure breach at Hugging Face, announcing plans to develop new reporting standards for misalignment.
The Seattle Times and Newsday have filed lawsuits against OpenAI and Microsoft over the unauthorized use of their content, claiming that AI systems are being trained on their proprietary articles, including those behind paywalls.
NVIDIA has released the beta version of PAIR, which allows AIinference processing to be distributed across multiple devices on a network. It enables tasks to be processed using the computing power of PCs, Macs, and DGX Spark without sending data to the cloud.