NerdNews

Meta is facing $1.4 trillion in damages in a social media addiction case brought by four states, alleging the company exploited young users for profit and misled consumers about addictive design features, potentially causing mental health problems in children.

China is considering limiting foreign access to its top AI systems, including open-source models, and restricting the use of US-made chips like Nvidia. This move is likely motivated by a desire to protect valuable Chinese IP and maintain a competitive edge in the global AI landscape.

Meta's new Muse Image model generates images using advanced reasoning and can be prompted with Instagram accounts. It's available in the US through the Meta AI app, Instagram, and WhatsApp, and will power over 30 new effects for Instagram Stories.

Meta has unveiled Muse Image, a new AI image generator built by Meta Superintelligence Labs. The feature allows users to create images, manipulate existing ones, and even use public Instagram photos to generate new images. However, this has raised concerns over privacy and data usage. Muse is available for free through the Meta AI app, Instagram Stories, and WhatsApp, with potential applications in advertising, interior decorating, and more.

Anthropic's Claude Cowork, a coding agent for general knowledge work, is now available on web and mobile for Max subscribers. The expansion allows users to start tasks on one device and pick up where they left off on another, making it feel like an administrative coworker. This move signals the coding agent wars are spilling into the rest of the office, with AI firms pushing their products beyond chatbots into everyday work surfaces.

The Model Context Protocol team has promoted its Enterprise-Managed Authorisation extension to stable status, adding a centralized way for organizations to control access to MCP servers through their identity provider. This extension aims to replace per-server consent prompts with a zero-touch flow, allowing users to sign in once and access approved servers without further setup.

A new attack pattern called Cordyceps exploits GitHub Actions workflows, allowing attackers to gain access to repository secrets and execute malicious code. This vulnerability exists because of how workflows connect, making it difficult for security scanners to detect. Immediate fixes include preferring pull_request over pull_request_target and gating privileged workflows behind manual approval.

Microsoft is testing a new Cloud Rebuild feature in Windows 11 Insider Preview builds, allowing for remote system reinstalls from the cloud for devices with persistent problems. The feature downloads the target Windows image and device drivers from Windows Update, restoring the device to a clean state without requiring USB media or a custom image.

Momentic rearchitected its caching system by transitioning from PostgreSQL to ClickHouse, handling 2 million queries per day with 250 ms latency. ClickHouse's column-oriented database and sparse primary indexes improved efficiency at scale.

Aaron Erickson explains how NVIDIA designs and tests purpose-built AI agent hierarchies, outlining the importance of balancing deterministic tools with agentic discovery to build reliable AI systems at scale.

DeepSeek, a Chinese AI company, is reportedly developing its own AI chips for inference, aiming to reduce reliance on third-party providers like Huawei and NVIDIA. The company is hiring engineers and speaking to manufacturers, signaling a potential new entrant in the competitive Chinese AI market.

Elon Musk's AI company xAI has been officially rebranded as SpaceXAI, five months after merging with SpaceX. The new branding and logo were announced on X, with plans to build orbital data centers and create a space-based data center using a million satellites.

Savi Security's app uses AI to detect and prevent scams, including live-call monitoring. The company was founded by brothers Patrick and Ryan Coughlin after their mother was targeted by a convincing AI-generated kidnapping scam. The app is available for iPhone and Android, and charges $8/month or $63/year to cover an entire family.

Meta's new Muse Image model can generate images based on user prompts and incorporate other Instagram users' likenesses. The model is part of the growing Muse family of AI models and will power image-making tools across Meta's apps, including Instagram and WhatsApp.

AWS has expanded its DevOps Agent with AI-powered release management capabilities, including Release Readiness Review and Autonomous Release Testing, to validate code changes before production. This move aims to bridge the gap between AI-accelerated code creation and traditional review and testing processes, enabling engineering teams to evaluate production readiness and enforce organizational standards.

A new research paper found that AI agents can consume up to 136.5 times more energy per query than generative AI models, with an average of 348.41 watt-hours per query, and may lead to significant energy demand if widely adopted.

SpaceXAI is reportedly launching a new frontier AI model, built jointly with AI startup Cursor, which SpaceX is acquiring for $60 billion. The model is being compared internally to Claude Opus 4.8 and OpenAI's GPT 5.5.

Tesla's latest iOS app update includes code suggesting the company may use its in-car camera to verify a driver's identity before allowing the use of its Full Self-Driving feature. This potential move could be related to laws requiring age verification and protecting minors from harmful content. However, it may face resistance from drivers who are uncomfortable with ID checks and handing over control of their car's features.

Meta has developed a tool to detect images and videos created with its new image generation model, Muse Image. The tool uses a watermarking system called Content Seal, which remains in place even when the image is cropped, compressed, or resized. The detection tool is currently limited to images created with Muse Image, but Meta plans to expand it to AI-generated and edited videos as well.

Forterra's autonomous ground vehicles have been deployed in Ukraine for 9 months, completing over 1,100 missions and carrying 777,440 pounds of cargo. The vehicles have proven useful but have limitations, such as requiring teleoperation in combat zones and being too valuable to be deployed freely.