Desert Ant Labs Launches Specialized On-Device AI Models and SDKs
A new European AI lab, Desert Ant Labs, has released its first 18 specialized models designed for fast, efficient on-device intelligence across audio, vision, and text.
Latest
Teaser Trailer Released for 'Artificial', a Film Centered on the Events Surrounding Sam Altman
A teaser trailer has been released for the film 'Artificial', which focuses on the events of 2023 when OpenAI founder Sam Altman was removed by the board of directors, only to return in less than two weeks.
DeepSeek Revises API Pricing, Lowering Input Cache Hit Rates for Flash Series
DeepSeek has announced a revision of its Flash series API pricing, effective 2026-09-10 at 12:00 Beijing Time. The unit price for input cache hits during off-peak hours will be reduced to $0.003.
OpenAI Sued Over Claims That ChatGPT Encouraged User's Delusions
A California man has sued OpenAI, claiming that interactions with ChatGPT deepened his religious delusions and caused emotional distress leading to self-harm.
OpenAI Releases Next-Gen GPT-6 Astra, Delivering High Task Execution Capability and Cost Efficiency
OpenAI has announced the release of GPT-6 Astra, a next-generation model designed for deep integration into business processes, now available via ChatGPT Work, Codex, and API.
Anthropic Researcher Warns There Is More Than a 10% Chance AI Could Cause Human Extinction
AI safety researcher Evan Hubinger stated that there is more than a 10% chance that AI will cause human extinction within the next 10 years, expressing concerns that technological evolution could pose an existential risk to humanity.
Anthropic Safety Researchers Express Concern that AI Could Cause Human Extinction Within 10 Years with Over 10% Probability
Safety researchers at Anthropic have indicated that there is a more than 10% chance that AI could lead to the death of all humanity. Former employees have also warned against a development race that neglects safety.
vLLM v0.29.0 Released — Model Runner V2 Now Default for All Models, New Model Support and Performance Gains
vLLM has released v0.29.0, which adopts Model Runner V2 as the default for all models and adds support for new models such as Hy4-preview and Qwen3.8-Flash-Next.
Former Anthropic Researcher Leaves Company, Warning of AI Existential Risk to Humanity
Jacob Coxon, a researcher who previously worked at both OpenAI and Anthropic, has revealed his departure from the company, warning of the dangers that the rapid advancement of AI technology poses to humanity.
MIT Successfully Automates Qubit Measurement Using GPT-5.6 Sol
MIT's EQuS group has demonstrated that an AI agent powered by GPT-5.6 Sol can autonomously execute measurement workflows for superconducting qubits.
OpenAI Considering Temporary Halt of New Pro Plan Sign-ups Due to Surging Demand for New Model
OpenAI Product Lead Thibaut Soultiao stated that the company may temporarily suspend new registrations for the ChatGPT Pro plan due to expanding demand for the new GPT-6 Astra model.
Xiaomi Begins Beta Testing for AI Desktop Agent Xiaomi MiMo Desktop
Xiaomi has launched beta testing for Xiaomi MiMo Desktop, an AI desktop agent designed to handle administrative work and creative tasks. Participation is by invitation, and applications are now open on the official website.
Massive volume of child sexual abuse material ads found on Meta SNS, including AI-generated images of real children
The non-profit Tech Transparency Project (TTP) reported that over 300 advertisements containing child sexual abuse material (CSAM) were published on Facebook and Instagram in 2026 alone.
Meta Announces Muse, a Personal AI Agent for Travel Booking and Shopping
Meta has unveiled Muse, a personal AI agent capable of handling tasks such as sending emails, booking travel, and shopping on behalf of users, featuring dedicated virtual machines and a monitoring AI to ensure security.
Google-affiliated Organization Reports AI-driven Acceleration and Expansion of Cyberattacks
Google's Google Threat Intelligence Group reported that the misuse of AI is accelerating and scaling attacks, with some instances seeing thousands of credentials compromised in less than six hours.
The Discovery Problem: Users' Inability to Perceive AI Capabilities Becomes a Bottleneck for Adoption
Despite the high potential of AI, the "discovery problem"—where users cannot find specific ways to utilize the technology—is becoming a barrier to AI adoption.
DaVinci Resolve 21.1 updated, supporting integration with AI assistants such as Claude and ChatGPT
Blackmagic Design has announced DaVinci Resolve 21.1, a major update to its video editing software. It includes features such as workflow automation through integration with AI assistants and expanded capabilities for the Photo Page.
OpenAI Announces Discovery of Solution to Navier-Stokes Equations via Multi-AI Agent System
OpenAI announced that a multi-AI agent system utilizing internal models discovered a solution to the Navier-Stokes equations, a long-standing unsolved problem in mathematics. Meanwhile, suspicions have emerged regarding the potential misappropriation of other researchers' work.
OpenCode v1.18.30 Released — Adds System Prompts for GPT-6 and Fixes Various Providers
OpenCode has released v1.18.30, featuring the addition of Astra system prompts for GPT-6 models and fixes for providers including Bedrock, OpenAI, and Azure.
Inception Announces Diffusion-Based Mercury 2.5, Enabling High-Speed Generation at 1,107 Tokens Per Second
Inception has unveiled Mercury 2.5, a new model employing a diffusion model that generates and refines multiple tokens in parallel, enabling faster output compared to traditional autoregressive LLMs.