Google researchers have proposed Procedural Graph, which manages an agent's next actions as "procedures." By automatically updating the execution structure from histories of success and failure, it achieves performance that surpasses memory-based methods.
Microsoft has agreed to legally binding AI safety and privacy standards with the American Federation of Teachers (AFT) and others, ensuring that student and educator data is not used for AI training.
OpenAI has announced the appointment of Paul Christiano, founder of the Alignment Research Center, to the board of the OpenAI Foundation.
Jacob Coxon, a researcher at Anthropic, has left the company and issued a warning regarding the existential risks posed by superintelligent AI capable of self-improvement. The company's head of safety research has expressed agreement with these concerns.
Google DeepMind has released AlphaGenome Atlas, a platform that predicts the biological impact of approximately 9 billion single-nucleotide variant patterns in the human genome.
Risks are being pointed out regarding Prolific AI Psychosis, a state where AI increases the volume of output without accompanying an increase in substantial value.
Job postings and interviews with sources reveal that AI developer Anthropic is building an extensive system to monitor activists who oppose the rapid development of AI.
Google announced the introduction of live feeds to track American football games in real-time and AI-powered support features for Fantasy Football within Google Search.
Google DeepMind and Primordial Soup have produced a short documentary titled "Love, Rendered," which uses AI technology to recreate past memories that were never captured on film.
A new classifier has been released to determine whether code comments were written by a human or generated by AI. It operates locally in the browser and allows users to examine the features driving the determination.
IBM has released Granite Time Series PatchTST-FM-r2, a time series foundation model with high zero-shot forecasting performance, provided under Apache 2.0 and OpenMDW 1.0 licenses.
Amazon Web Services (AWS) has released the Nx Plugin for AWS 1.0, an open-source tool that uses AI to automatically generate infrastructure configurations and basic application code.
Frigade has announced the "Assist API," which allows AI agents to learn a product's UI and provide real-time operational guidance to users.
Geiger, an open-source tool capable of inspecting the configuration and permissions of AI agents running on machines, MCP (Model Context Protocol) servers, plugins, and AI extensions, has been released.
OpenAI has announced its new model, GPT-6 Astra, which delivers state-of-the-art performance in computer use, mathematics, and cybersecurity. Compared to the previous GPT-5.6 Sol, it achieves faster task completion times and higher benchmark scores.
The AI agent Hermes Agent has introduced a Memory feature that allows it to remember user preferences and work environments over the long term, maintaining past instructions and context even when starting a new chat.
OpenAI has stated that as AI capabilities advance, there is an urgent need for a policy framework to ensure safety. The company has also expressed its support for four bills passed in California that are awaiting the governor's signature.
Amazon Prime Video has announced a new feature that combines AI and visual effects (VFX) technology to synchronize actors' lip movements with translated audio.
Multiple development tools, including the open-source office suite LibreOffice, have been discovered within the cache folder of the ChatGPT desktop application.
According to 2025 data from the PISA study conducted by the OECD, students who use AI tend to have lower scores in fields such as science compared to those who do not.