According to a survey conducted by Change Holdings and e-Guardian, approximately 6 of business tasks are reported to be "human-led." A significant gap in the acceptance of AI-driven decision-making was particularly observed between the 20 and 50 generations.
A research team from Delft University of Technology in the Netherlands and other institutions announced that their investigation into citations of Beatles songs and lyrics in scientific paper titles revealed numerous instances.
OpenAI has announced the release of GPT-Live-1, a voice model for its API that enables simultaneous listening and speaking. Developers can freely select the backend inference models to build applications with natural voice experiences.
AI developer Anthropic has published its latest threat report summarizing instances where Claude was misused, reporting that AI is increasing the efficiency of attacks ranging from cyber operations to surveillance and information warfare.
OpenAI has introduced the "Data agent" to ChatGPT Work, allowing users to connect to corporate data sources to retrieve answers through queries and create interactive dashboards.
On September 11, Sakana AI released Fugu Ultra v2 and Fugu Max as the latest versions of Sakana Fugu, a system that combines multiple AI models. These are provided via an OpenAI-compatible API.
OpenClaw has released v2026.9.4, featuring an automatic rollback function for failed compatible updates, integrated plugin management, and cloud session preparation capabilities.
The J-HRTI Kanto Data Factory, designed to mass-produce training data for robots, has begun operations in Narashino City, Chiba Prefecture. Operated by a consortium of five private companies, the facility collects and shares motion data for humanoid robots.
OpenAI has released the GPT-Live-1 API, a voice model that allows fine-grained control over the speaking style and behavior of voice agents. Utilizing a full-duplex system, it supports interruptions and the delegation of inference functions.
OpenAI has released the public beta of the Agents API to support the creation of AI agents, providing access to the same harness and infrastructure as Codex via an API.
LLM Visualizer has been released, allowing users to experience the operating principles of LLMs—from raw text to output generation—through mathematics and code.
NEC has established the Corporate AI Workforce Division as of August 1, where autonomous AI agents take on various roles. In internal proof-of-concept tests, the company reported a reduction in work hours to approximately one-seventh of previous levels.
China's DeepSeek released the open model DeepSeek-V4.1-Flash on 2026-09-10, which supports image input. It has demonstrated performance exceeding Claude Opus 5 and GPT-5.6 Sol in multiple tests, while maintaining low API pricing.
Google has released the Gemini desktop app for Windows 10 and later, including ARM64 support. The app enables quick launching via shortcut keys and seamless integration with Google apps.
OpenAI has announced ChatGPT for Financial Services, combining the inference capabilities of GPT-6 Astra with specialized financial data.
Amazon has announced the introduction of "lip-sync" technology to Prime Video, utilizing AI and VFX to align the mouth movements of actors in dubbed content with the spoken dialogue.
Amazon Ads has announced a partnership with OpenAI to launch advertising on ChatGPT, starting with a pilot program for select advertisers in the United States.