Latest AI News

A groundbreaking study by Mount Sinai and Rabin Medical Center researchers shows that even the most advanced AI models, including ChatGPT, make surprisingly basic errors when navigating medical ethics scenarios. Published July 24, 2025, in npj Digital Medicine, the research reveals that AI systems often default to familiar but incorrect responses when presented with slightly modified ethical dilemmas, raising serious concerns about their reliability in healthcare settings. The findings underscore the critical need for human oversight when deploying AI in medical decision-making.

The Browser Company's Dia and Perplexity's Comet have introduced skill galleries that transform how users interact with AI-powered browsers. These galleries allow users to save prompts for repeatable tasks like code generation or event discovery, with options to explore community-created skills. According to a16z partner Olivia Moore, this represents a pivotal shift in web interaction by integrating AI directly into browsing workflows rather than requiring separate interfaces.

Independent MP Kate Chaney has introduced groundbreaking legislation in the Australian Parliament to criminalize AI technology designed to create child sexual abuse material. The Criminal Code Amendment Bill would make it an offense to download, possess, or distribute AI tools specifically developed to produce child sexual abuse content or to collect data for training such systems. The legislation, which follows a roundtable on AI-facilitated child exploitation, aims to address what Chaney describes as "a clear gap" in Australia's criminal code.

China has proposed establishing a global AI cooperation organization during the 2025 World Artificial Intelligence Conference in Shanghai. Premier Li Qiang announced the initiative on July 26, positioning it as China's response to calls from the Global South for more equitable AI development. The organization, likely to be headquartered in Shanghai, aims to promote international cooperation on AI technology development and regulation while bridging the digital divide.

Australian scientists at the University of Sydney, led by Professor David Reilly, have achieved a groundbreaking advancement in quantum computing with a CMOS-spin qubit chip that operates at ultra-low temperatures with minimal power. This innovation solves a critical bottleneck in quantum computer scaling by integrating control electronics directly with qubits, maintaining quantum coherence despite proximity. The breakthrough paves the way for silicon-based quantum computers with millions of qubits, potentially revolutionizing AI processing capabilities.

OpenAI is preparing to release GPT-5 in August 2025, according to multiple reliable sources including the AI Unraveled podcast and recent statements from CEO Sam Altman. This next-generation model will unify OpenAI's various AI technologies, combining traditional language processing with advanced reasoning capabilities. The release represents a significant milestone in OpenAI's roadmap, following the earlier launch of GPT-4.5 and specialized reasoning models.

McKinsey's groundbreaking 'Superagency in the Workplace' report reveals that while 92% of companies plan to increase AI investments, only 1% have reached implementation maturity. The research, shared by McKinsey experts Michael Chui, Roger Roberts, and Lareina Yee on July 29, 2025, highlights how AI differs fundamentally from previous breakthrough technologies by offering capabilities beyond information access, including reasoning, dialogue, and decision-making. The study finds that employees are more ready to embrace AI than leaders realize, with leadership alignment—not technology or employee willingness—being the primary barrier to successful implementation.

Sydney-based Archistar has officially launched its AI-powered eCheck platform in partnership with Los Angeles authorities to accelerate rebuilding after devastating January wildfires. The technology, which automatically validates building designs against local codes, represents California's first large-scale deployment of AI in disaster recovery. This initiative follows Archistar's recent strategic partnership with the International Code Council, positioning the platform as a trusted global solution for modernizing building approvals.

UC Riverside researchers and Google have developed UNITE, a groundbreaking AI system that detects deepfakes even when faces aren't visible in videos. Unlike traditional detection methods, UNITE analyzes entire video frames, including backgrounds and motion patterns, to identify synthetic or manipulated content. This universal detector represents a significant advancement in combating increasingly sophisticated AI-generated videos that threaten information integrity.

Los Angeles-based Steadfast Robotics is transforming post-wildfire reconstruction in Southern California with AI-driven robots and software that reduce site preparation time by 60%. Founded in 2023 by engineer Elena Vasquez, the company combines autonomous machines with regulatory compliance software, partnering with Australian firm Archistar to integrate eCheck AI for automating zoning and permitting processes. This technology creates a seamless loop from assessment to approval, addressing a critical bottleneck in disaster recovery.

Microsoft launched Copilot Mode for its Edge browser on July 28, 2025, turning the traditional browsing experience into an AI-assisted journey. This experimental feature combines search, chat, and navigation into a unified interface that can understand users' research context across multiple tabs and predict their next actions. Currently free for Windows and Mac users with access to Copilot, the feature positions Microsoft competitively in the rapidly evolving AI browser landscape.

OpenAI and Oracle have announced a significant expansion of their Stargate AI infrastructure project, adding 4.5 gigawatts of data center capacity in the United States. This development, which will create over 100,000 jobs across construction, operations, and service industries, advances their $500 billion commitment to build 10GW of AI infrastructure nationwide. The Stargate I facility in Abilene, Texas is already partially operational, with Oracle delivering advanced Nvidia GB200 GPU racks to power next-generation AI research.

Adobe has launched Harmonize, a groundbreaking Firefly-powered AI feature for Photoshop that automatically adjusts color, lighting, shadows, and visual tone to create seamless photo composites with just a few clicks. First previewed as Project Perfect Blend at Adobe MAX 2024, Harmonize is now available in beta across desktop and web versions of Photoshop, with early access on iOS. This technology significantly reduces the need for time-consuming manual adjustments, democratizing professional-level editing for creators of all skill levels.

Yelp is testing innovative AI-stitched videos that automatically compile user-submitted photos, videos, and business information into engaging visual previews of local restaurants. These dynamic videos showcase the dining experience, highlight popular dishes, and provide a more immersive preview than static images alone. The feature represents Yelp's ongoing effort to enhance discovery and connection with local businesses through AI-powered innovations.

Google has made its advanced Veo 3 AI video generation model available to select Workspace customers as of July 29, 2025, following its initial unveiling at Google I/O in May. The technology, which creates high-quality videos with synchronized audio including dialogue and ambient sounds, is rolling out first to Rapid Release domains with Scheduled Release domains to follow in August. Workspace users can access Veo 3 through the Gemini app and Google Vids, with usage limits of 3 daily generations for most enterprise users.

Harmonic, co-founded by Robinhood CEO Vlad Tenev, has launched Aristotle, an AI chatbot app that promises hallucination-free mathematical reasoning. The model, which achieved gold medal performance on the 2025 International Math Olympiad, uses the Lean programming language to formally verify its answers. Harmonic aims to create what it calls 'mathematical superintelligence' (MSI) to revolutionize fields relying on mathematical precision.

Google has unveiled new educational capabilities in AI Mode for Search, powered by Gemini 2.5, their most intelligent model to date. These features are specifically designed to help learners, educators, and curious minds explore complex topics through advanced reasoning and multimodality. The update includes PDF upload support, interactive study tools, and deeper research capabilities that transform how users can interact with educational content.

The Trump administration's recently unveiled AI Action Plan aims to accelerate data center development in Texas by streamlining federal permitting and reducing environmental regulations. Signed on July 23, the plan includes over 90 federal policy actions across three pillars designed to maintain U.S. dominance in artificial intelligence. While promising economic growth and technological advancement, the initiative has sparked concerns about water usage and environmental impacts in drought-prone regions.

Google is providing more than 30 AI tools to educators through Google Classroom at no cost, enabling teachers to generate content and resources with Gemini. The comprehensive suite helps educators kickstart lessons, brainstorm ideas, and create differentiated learning materials for students. In the coming months, Google will launch teacher-enabled interactive AI experiences including Teacher-led NotebookLM for creating study guides and Audio Overviews, and Teacher-led Gems that act as AI experts to support students.

China Southern Power Grid (CSG) showcased its cutting-edge AI innovations at the 2025 World Artificial Intelligence Conference in Shanghai from July 26-28. As a strategic partner of the event, CSG presented its exhibition themed 'Empowering Every kwh with AI' and hosted a forum on AI-driven innovation in the power sector. The company demonstrated various AI-powered applications highlighting its initiatives to integrate artificial intelligence into energy systems and support the development of a smarter industrial ecosystem.