Moonshot's 2.8 Trillion Parameter Kimi-K3

Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model with a 1 million-token context window that matches or exceeds leading proprietary U.S. models on several coding, reasoning, and agent benchmarks while remaining openly available.

DON Strategy to Weaponize Data and AI

The Department of the Navy's new strategy directs the Navy and USMC to treat data as a strategic asset, accelerate operational AI adoption, modernize digital infrastructure, and strengthen workforce AI literacy to improve warfighting readiness and decision-making.

Thinking Machines Releases Inkling Model

U.S.-based Thinking Machines released Inkling, a 975B-parameter open-weight multimodal model, giving enterprises and developers a domestic alternative to leading Chinese open-weight models while emphasizing flexibility over benchmark leadership.

DoorDash Launches Command-Line AI Tool

DoorDash has released a limited beta of a command-line interface (CLI) tool that allows AI agents to search for stores, find deals, and place orders. The tool is currently available to macOS developers in the U.S. and Canada via a waitlist.

Meta Pulls Muse Image AI After Backlash

Meta removed its Muse Image AI feature from Instagram and its AI app three days after launch. The tool allowed users to generate images using public Instagram photos as references without requiring explicit user opt-in.

Humanoid AI's Wheeled Robot Strategy

UK-based Humanoid AI is prioritizing a wheeled humanoid robot, the HMND 01, for industrial use. The platform utilizes a four-layer AI architecture and is designed to meet existing CE certification standards for rapid deployment.

George Lucas Endorses AI in Filmmaking

George Lucas said AI will make filmmaking easier and is an inevitable technological progression, arguing that filmmakers should embrace the technology while cautioning that humans remain responsible for how it is used.

White House Launches Gold Eagle

The White House launched Gold Eagle, an AI-driven coordination mechanism designed to accelerate the detection, prioritization, and remediation of software vulnerabilities across critical infrastructure.

DARPA Flies AI-Controlled F-16

DARPA and the U.S. Air Force successfully conducted in-air testing of an F-16 modified with the VENOM Autonomy Kit, allowing an AI agent to autonomously control the aircraft while a human pilot monitors the system.

AI Accelerates Battery Material Design

LLNL researchers integrated molecular dynamics simulations with physics-informed machine learning to model atomic-scale interactions in sodium-ion and lithium-ion battery materials.

Publishers Sue Google Over AI Training

A group of publishers and authors filed a class action lawsuit against Google, alleging the company used copyrighted works to train its Gemini AI models without authorization.

ANL Advances Autonomous Discovery

Argonne is expanding AI-driven autonomous laboratories that combine robotics and machine learning to design, execute, and refine experiments, accelerating research across biology, chemistry, and materials science.

AI agents create virtual playgrounds

Robots walking down the street, surrounded by astounded onlookers, is an increasingly common sight. But these machines aren’t yet the do-it-all assistants you’d want working in a kitchen or factory, and a major bottleneck is data.

Benchmark Tests Humanoid Robot Force

Researchers developed ThorArena, a simulation-based benchmark that evaluates how humanoid robots handle physical forces by mapping human motion and force data onto virtual models.

AI Speeds Brain-Computer Interface Learning

Carnegie Mellon researchers combined AI with touch-based feedback to help first-time users learn to control a noninvasive brain-computer interface faster and more accurately, achieving up to 86% accuracy while improving long-term skill retention.

BYD Expands Into Humanoid Robotics

Automotive manufacturer BYD is transitioning from using third-party humanoid robots in its factories to developing its own proprietary models for industrial, commercial, and eventually domestic use.

Google DeepMind Bioresilience Strategy

Google DeepMind and Isomorphic Labs have launched a bioresilience program to prevent AI misuse and assist government and research partners in detecting and responding to biological threats.

PrismML Releases Bonsai 27B Models

PrismML released 1-bit and ternary weight versions of the Qwen3.6-27B model, enabling 27B-parameter class performance on mobile devices and laptops. The models are available under an Apache 2.0 license.

Grok Build Data Privacy Incident

Security researchers discovered that the Grok Build CLI tool was automatically uploading entire user repositories and Git histories to cloud storage. SpaceXAI has since implemented a server-side fix to stop the unauthorized data transfers.

NotebookLM is now Gemini Notebook

Google renamed NotebookLM to Gemini Notebook, aligning it with the Gemini ecosystem while preserving it as a standalone research tool and expanding integration with Gemini

Secure Sandboxes for AI Agents

Perplexity introduced SPACE, a sandbox that isolates every AI agent task in ephemeral microVMs with hardware-backed virtualization, fine-grained permissions, proxy credentials, and network controls.

Roblox Launches Mobile AI Creation Tools

Roblox introduced Build, an AI-powered mobile tool that lets users create playable games from text prompts, along with AI agents to help creators test, analyze, and improve their games.

NVIDIA Launches Jetson Thor T3000/T2000

NVIDIA introduced the T3000 and T2000 modules based on the Blackwell architecture, designed to provide scalable, power-efficient AI compute for robotics and edge applications.

OpenAI and Google AI Sold to Chinese Firms

OpenAI and Google confirmed providing AI services to subsidiaries of Chinese tech giants Alibaba, Baidu, and Tencent. These firms are currently on the Pentagon's blacklist due to alleged ties with the Chinese military.