Artificial Intelligence News Live: Google enhances Home app with expanded Gemini AI smart home controls for camera search

Google is integrating its Gemini AI more deeply into its smart home ecosystem, providing broader controls within the Home app, according to a report in The Verge. Users can now employ the Gemini app to search their camera clip history by typing specific queries, and can also broadcast their voice through connected speakers using Gemini. This development signifies a continued effort by Google to transition smart home tasks, including more advanced AI-powered capabilities, from the Google Assistant to the Gemini platform, though voice dictation for these new Gemini-based camera search queries is not currently available.

A new framework called AlphaOne has been introduced, presenting an approach to modulate the "thinking" processes of Large Language Models (LLMs), VentureBeat says in a report. This method aims to improve both the accuracy and efficiency of these AI models. A key claimed advantage of AlphaOne is its potential to achieve these performance enhancements without the need for expensive and time-consuming retraining of the core models, offering AI developers a new way to control LLM behavior.

Amazon's generative AI-powered voice assistant, Alexa Plus, is now in an Early Access program with over one million users, an increase from "hundreds of thousands" in May 2025, as it rolls out at an "increasing pace" with broader availability expected over the summer, The Verge said in a report. Initially announced in September 2023 and relaunched in February 2025 under Panos Panay, head of Amazon devices and services, the assistant aims for a careful rollout. While Amazon spokesperson Eric Sveum states nearly 90% of announced features are live, including enhanced conversational abilities and calendar management, several key functionalities like browser access and certain ordering capabilities are still pending for "coming weeks and months." User feedback is generally positive regarding its conversational skills, though some bugs and issues with smart home device compatibility compared to the previous Alexa version have been reported by early users.

OpenAI has launched o3-pro, an artificial intelligence model that the company claims is its most capable reasoning model yet, TechCrunch said in a report. This new model is described as an enhanced version of OpenAI's existing o3 AI reasoning model, signifying continued development in advanced AI capabilities.

Director of National Intelligence Avril Haines said Tuesday that artificial intelligence is accelerating the work of America's intelligence services, offering potential to save money and allow officers to focus on information gathering and analysis, AP says in a report. Haines highlighted that responsibly used AI programs can manage tasks like human resources and are speeding up intelligence work, including processes involved with historical records such as the JFK assassination files. This application of AI aims to address challenges related to the pace of intelligence operations.

Academic publishers are actively signing licensing deals with artificial intelligence companies to establish new revenue streams, a move driven by concerns over diminishing financial outlooks due to US research funding cuts, Bloomberg said in a report. These agreements grant AI firms access to extensive, high-quality, peer-reviewed academic content, which is crucial for training their models, especially those focused on scientific, technical, and medical domains. This development is part of a wider trend where various content creators are licensing their material to AI companies, occurring amidst ongoing legal debates surrounding copyright and the use of protected works for AI training.

Qualcomm has shared its vision for integrating generative AI directly onto smart glasses, a move aimed at enabling unique conversational experiences through the eyewear, VentureBeat said in a report. An executive from the company described interactions with these AI-powered smart glasses as potentially "strange and most interesting." This focus on "on-glass Gen AI" suggests that significant processing will occur locally on the device, likely using Qualcomm's specialized chips. The announcement emphasizes a technology demonstration and future roadmap, not an immediate consumer product release.

OpenAI intends to utilize Google Cloud's infrastructure from Alphabet to address its escalating needs for computational resources, a significant development given the intense competition between OpenAI and Google in the artificial intelligence field, Reuters said in a report. This move, reportedly based on information from three sources, highlights the immense demand for AI compute power, leading to such collaborations even between direct competitors. While OpenAI has an established deep partnership with Microsoft Azure, this potential agreement with Google Cloud suggests a strategy to diversify and secure additional capacity for training and running its large language models.

Enterprise AI startup Glean announced it has secured $150 million in a Series F funding round led by Wellington Management, resulting in a post-money valuation of $7.2 billion, TechCrunch said in a report. Glean's primary focus is on providing AI-powered solutions specifically to businesses and organizations. This Series F round indicates Glean is a relatively mature startup that has undergone multiple prior rounds of significant investment.

Nvidia's stock experienced a rise following the announcement that the company will supply AI accelerators to Saudi Arabia's Humain, Investor's Business Daily said in a report. Humain, identified as a new AI entity, is in the process of constructing a substantial 500-megawatt AI data center.

OpenAI experienced a partial outage on Tuesday morning affecting its popular services ChatGPT, Sora, and its API, according to a report in TechCrunch. The company acknowledged the issues on its status page, stating it began investigating late on Monday night and that the disruption persisted through Tuesday morning, with an update around 5:30 am PT on Tuesday indicating ongoing problems.

Mistral has released Magistral, its initial family of artificial intelligence reasoning models, TechCrunch said in a report. These models are designed to work through problems step-by-step for improved consistency and reliability in complex subjects such as math and physics, positioning them similarly to other advanced models like OpenAI's GPT-4o (released May 2024) and Google's Gemini 1.5 Pro (updated earlier in 2024). The report indicates 'Magistral' refers to a family of models, suggesting potential variations.

OpenAI's ChatGPT service suffered a significant and widespread outage that persisted for several hours, impacting numerous users and developers who rely on the AI tool, ZDNet said in a report. OpenAI officially acknowledged that its service was experiencing issues, confirming it was aware of the widespread problem and presumably working on a resolution. Such disruptions underscore the increasing dependence on AI tools across various sectors, making service interruptions highly impactful for both individual users and businesses that integrate ChatGPT for tasks like content generation, coding, and customer support, as well as for applications leveraging its API.

As artificial intelligence agents become more potent and their behavior less predictable, Cisco has unveiled a strategy and accompanying tools aimed at securing internal networks against potential rogue AI agent activities, ZDNet said in a report. The company's approach focuses on enhancing network visibility, enabling the tracking of AI agent behavior, and implementing measures to contain or prevent disruptive actions before they can propagate throughout an organization's infrastructure.

As Tesla prepares to launch its robotaxi service in Austin, Texas, within days, neither the US federal government nor the City of Austin has publicly detailed how the company's vehicles will utilize teleoperations, described as self-driving’s critical safety feature, Wired says in a report. This lack of transparency regarding remote human assistance for Tesla's autonomous vehicles in Austin raises questions, particularly as such systems are crucial for handling complex scenarios and are often more openly discussed by other autonomous vehicle operators.

Microsoft is testing a new AI-powered "enhanced search" feature for its Edge browser's history, which started appearing in beta versions last week, The Verge says in a report. This optional feature enables users to find previously visited websites even with imprecise queries like phrases, synonyms, or typos, utilizing an on-device model trained with user data that Microsoft states "never leaves your device and is never sent to Microsoft." The article notes this approach presents less of a privacy risk compared to Microsoft's recently criticized Recall feature. Alongside this AI history search, Microsoft is also introducing a media control center to Edge.

Zip has launched 50 specialized artificial intelligence agents aimed at automating enterprise procurement, handling tasks such as contract reviews and compliance checks, VentureBeat said in a report. Technology companies OpenAI and Canva are among the early adopters of this new AI solution, through which Zip is targeting $4.4 billion in potential savings for clients by enhancing operational efficiency.

Czech online grocer Rohlik Group is spinning out its AI-driven platform designed for rapid grocery deliveries, with ambitions to offer the technology worldwide and compete with established players such as Ocado Group Plc, Bloomberg said in a report. This strategic move involves creating a distinct business unit for the platform, which utilizes artificial intelligence to optimize various logistics aspects, including demand forecasting, inventory management, warehouse operations, and last-mile delivery routing, thereby enhancing efficiency for grocery retailers.

Human line judges will be absent from Wimbledon this year as the tournament transitions to a fully automated electronic line calling (ELC) system, which employs artificial intelligence and 12 cameras courtside to make line calls, Bloomberg said in a report. This system is provided by Hawk-Eye Innovations, a Sony subsidiary, reflecting a broader trend in major tennis tournaments moving away from human line judges to enhance officiating accuracy and streamline operations.

The Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) has announced a tentative contract with several video game companies, potentially resolving an almost year-long strike centered on the use of artificial intelligence, AP said in a report. The strike, which began around July 2023, was initiated after negotiations stalled over protections for performers against the unregulated use of AI, which the union described as a significant threat to their work in the video game industry. This tentative agreement, which still requires ratification by SAG-AFTRA members, aims to address concerns about AI's impact on voice and performance capture work.

UK digital bank Starling has launched a new opt-in AI-powered tool, termed "enhanced search," enabling customers to analyze their spending habits by asking natural language questions, according to a report in The Verge. Built with Google Gemini, the feature suggests personalized prompts based on spending patterns, automatically categorizes transactions into over 50 customizable categories, and presents data through graphs and analytics to help users understand their finances. Harriet Rees, CIO of Starling Bank, stated the aim is for users to engage with their money management in a natural way.

The British government announced on June 10, 2025, its intention to allow limited self-driving car pilots on UK roads, potentially starting as early as next year, Wired said in a report. This move signifies a forthcoming regulatory change or permit system, likely requiring autonomous vehicles to operate under specific conditions such as defined routes, particular speeds, or with a safety driver present. Companies including Uber and the British AV firm Wayve are reportedly gearing up to participate in these pilots, an initiative that builds on prior groundwork including the 2022 Law Commission reports which recommended a new legal framework for automated vehicles.

Researchers at The Hong Kong Polytechnic University (PolyU) have introduced VideoMind, a new video-language AI agent aimed at enhancing the ability of artificial intelligence models to comprehend long videos, AP says in a report. Current AI models often struggle with extended video content, and VideoMind seeks to address this by enabling AI to perform long video reasoning and question-answering tasks by mimicking human cognitive processes. The framework for VideoMind incorporates an innovative Chain-of-Low-Rank Adaptation (LoRA) strategy, which is expected to accelerate the development of generative AI-assisted video analysis.

Meta is establishing a new artificial intelligence lab with the ambitious goal of pursuing 'superintelligence', The New York Times said in a report. This initiative is part of a significant reorganization of Meta's AI efforts, overseen by chief executive Mark Zuckerberg, and is set to include Alexandr Wang, the founder of Scale AI.

A new paper from Apple researchers reveals significant challenges for advanced artificial intelligence, specifically large reasoning models (LRMs), which reportedly suffer a "complete accuracy collapse" when presented with highly complex problems, The Guardian says in a report. These findings point to "fundamental limitations" in current cutting-edge AI models and raise doubts within the technology industry about the feasibility of the ongoing race to develop systems that can match human intelligence across all tasks, particularly in complex reasoning scenarios.

Scammers are reportedly employing artificial intelligence to facilitate the fraudulent enrollment of fake students in online college courses, with the aim of stealing financial aid, AP said in a report. This scheme has resulted in public colleges losing millions of dollars and has burdened unsuspecting individuals, whose identities were stolen, with student debt for colleges they never attended, as illustrated by an incident involving a San Francisco resident, Heather Brady, whose identity was used for an application to Arizona Western College.

Sam Altman, CEO of OpenAI, the company behind ChatGPT, recently discussed key aspects of artificial intelligence, focusing on its safety considerations and practical utility, as per a report in CNN. In the interview, Altman also addressed the internal leadership challenges at OpenAI, specifically the boardroom events in November 2023 that resulted in his temporary removal and subsequent reinstatement as CEO.