NerdNews

Microsoft has introduced Copilot Vision, an AI-powered feature that captures and analyzes user screen activity, sending the data to Microsoft servers for processing. The feature is designed to provide personalized assistance and guidance, but raises concerns about user privacy and data security. Copilot Vision is part of Microsoft's broader push to integrate AI into Windows 11, which also includes features like agentic AI and improved natural language processing.

Google is rolling out generative AI video features to YouTube Shorts and Google Photos, allowing creators to generate new video elements from text prompts and photos. The features are currently based on the Veo 2 model, but an upgrade to Veo 3 is planned for later this summer.

Google's AI Overviews have reached 2 billion monthly users, while AI Mode has 100 million monthly active users in the US and India. Gemini app has grown to 450 million monthly active users, with over 9 million developers building with it.

YouTube Shorts is rolling out an image-to-video AI tool and new AI effects, allowing creators to turn static images into videos and add artistic effects to their content. The features are powered by Google's Veo 2 AI model and will be available in the US, Canada, Australia, and New Zealand initially.

Google's CEO Sundar Pichai stated that AI is positively impacting every part of the business, with features like AI Overviews and AI Mode performing well. The company reported revenue of $96.4 billion, up 14% year over year, and plans to increase its capital expenditures to $85 billion.

OpenAI's ChatGPT is rolling out a new 'personality' feature, allowing users to choose between multiple personalities, such as 'Robot', 'Cynic', 'Listener', and 'Sage', to personalize interactions. The feature is slowly rolling out to users worldwide.

Microsoft SharePoint has been attacked, compromising over 400 organizations, including the US Department of Energy. The attacks exploit two vulnerabilities, CVE-2025-53770 and CVE-2025-53771, which allow remote code execution and security bypass. Microsoft has released patches for the vulnerabilities, but the attacks are ongoing.

The K Prize, a new AI coding challenge, has announced its first winner with a score of 7.5%. The challenge tests models against real-world programming problems and is designed to be a contamination-free version of existing benchmarks. The low score highlights the need for better evaluation of AI models.

The Trump administration has published its AI Action Plan, which prioritizes growth and deregulation over caution and regulation. The plan aims to build out AI infrastructure, cut red tape for tech companies, and compete with China. It also focuses on protecting free speech and 'American values' in AI development, and includes provisions for AI safety and security.

Hackers have breached at least 400 organizations by exploiting a zero-day vulnerability in Microsoft SharePoint, with affected organizations including the National Nuclear Security Administration. The bug allows attackers to remotely run malicious code on affected servers.

Proton has launched Lumo, a privacy-respecting encrypted AI assistant that does not log user conversations and doesn't use their prompts for training. Lumo is based on open-source large language models and utilizes Proton's open-source encryption scheme.

OpenAI and partners are developing a 4.5 gigawatt AI data center in Texas, expanding the Stargate AI infrastructure platform to over 5 gigawatts of total capacity, which will power roughly 4.4 million American homes and support next-generation frontier research.

Google CEO Sundar Pichai expressed excitement about the company's partnership with OpenAI, providing cloud computing resources to train and serve AI models. This partnership marks a significant deal for Google Cloud, despite OpenAI being a major competitor to Google Search. The move is seen as a strategic play to grow Google Cloud's business, which has shown significant growth in the second quarter of 2025, reaching $13.6 billion in revenue.

Meta researchers are developing a wristband that uses surface electromyography (sEMG) to detect electrical signals from muscle activity, allowing users to control computers with hand gestures. The goal is to provide a less invasive tool for people with motor disabilities. The wristband is being tested with people who have spinal cord injuries and can interpret intended actions even before they are made.

Tesla plans to bring its robotaxi service to new markets, including cities in Florida, Nevada, Arizona, and California, with Elon Musk aiming for 50% of the US population to have access by the end of 2025. The company is testing its Full Self-Driving feature in Europe and China, despite current limitations and regulatory hurdles.

The UK government has announced partners in its £3.7bn connected and automated mobility sector drive, including Transport for London and Admiral. The initiative aims to create a £40bn sector and 6,500 jobs by 2040, with a focus on innovative technologies and services.

T-Mobile's Starlink satellite service is now available after months of testing, connecting smartphones in areas beyond traditional networks. The service uses over 650 low-Earth orbit satellites and supports texting on Android and iOS devices, with access to optimized apps like Google and WhatsApp. It's free for T-Mobile Experience Beyond subscribers, $10/month for others, and also available to AT&T and Verizon users.

Meta's body-reading wristband uses sensors to read electrical signals in muscles, allowing users to control devices with hand gestures. The latest prototype boasts improvements, including a neural network that quickly assesses signals and tailors the device to the user, making it more feasible for commercial use.

Meta's wristband uses electrical signals from the brain to predict and control device interactions with hand gestures, set to launch with Hypernova glasses.

President Trump announced a 'commonsense application' of intellectual property and AI, suggesting that AI companies won't be forced to pay for copyrighted material used in training models. He also doubled down on his anti-woke rhetoric, stating that the US will 'get rid of woke' in AI models. The remarks came during a keynote speech at a summit hosted by the All-In Podcast, following the release of the Trump administration's AI Action Plan, which outlines over 90 policy recommendations to ensure the US wins the 'AI race' against China.