Google Sets Guided Learning as K–12 Search Default in January

Google plans to make Guided Learning the default search experience for K–12 students in January, while keeping AI Mode and AI Overview available. Students could lose access to Google Search if their districts do not allow the AI-powered version. Parents and education advocates have objected, and Common Sense Media reported that in 180 tests Google’s AI Mode completed homework and sometimes produced fake or unreliable answers. California districts are weighing AI limits; Los Angeles Unified has already restricted generative AI.

Universities are also reconsidering how they teach and assess work produced with AI. A Norwegian government committee recommends more supervised, in-person exams and phasing out short take-home exams, while building lecturers’ AI skills and evaluating how AI affects courses and professional practice. Its report warns that AI can undermine learning without suitable teaching methods and offers students nine recommendations, including avoiding shortcuts and thinking independently.

OpenAI’s Codex hackathon at TwitchCon in San Diego prompted debate over AI and creative work. Some streamers welcomed tools for captions, speech-to-text and clipping, while others objected to content scraping and called for clear opt-outs. Twitch CEO Dan Clancy said the setting lets creators express whether Amazon may use their Twitch content for AI training, but he was not involved in Amazon’s plans. It remains unclear whether Amazon has used Twitch content to train models; Chief Product Officer Mike Minton declined to discuss past use, citing ongoing litigation.

Businesses in Northern Ireland are using AI for practical tasks. Newry-based Kukoon Rugs applies it in finance, marketing, supply chains and product design; CEO Paul Vallely says it speeds work, improves accuracy and frees staff from routine tasks. Seating Matters uses AI for back-office work such as accounting, finance and information gathering. Managing Director Martin Tierney says it is weak in customer service and urges staff to use it sensibly and with safeguards.

At SC26, Carahsoft and more than 50 partners will showcase high-performance computing, hardware, AI and education solutions through sessions and workshops on topics including data infrastructure and quantum innovation. The conference theme is “HPC Unites.” Caltech professor and Accelerated Understanding co-founder Animashree Anandkumar will deliver the keynote, “The Next Scaling Law after LLMS: AI that Understands the Physical World”; Carahsoft will be at Booth #2006.

A study shared by Vivek Subbiah of NYU Langone Health maps how physical work may affect AI exposure across 56 medical specialties. Dermatology, with 68% physical work, orthopedic surgery (53%) and ophthalmology (45%) were among the most shielded. Pathology (99%), diagnostic radiology (96%) and radiation oncology (62%) were among the most exposed. Across specialties, 24% of work involved physical tasks, 21% was nonphysical and potentially substitutable, and 55% involved Evaluation and Management services.

A separate look at Google’s Chubby Lock Service, described in a 2006 paper by Mike Burrows, highlights a design principle relevant to AI infrastructure: provide coordination as a shared service rather than making every application build its own. Chubby prioritized simplicity and reliable, low-volume storage over speed, using concepts including leases, sequencers and shared metadata. These ideas may inform reliable systems built from distributed pipelines, vector databases, inference gateways and GPU workers.

Researchers comparing human chess games with Stockfish–Leela games found that human games had greater average strategic tension after the first 20 moves, but people later simplified positions while AI games remained more complex. Tension also rose with human ratings, more thinking time and deeper Stockfish searches. The study did not establish that greater tension causes stronger play.

John Provenzano, CEO of The 9-1-1 Association, argues that AI can make emergency communications faster and more reliable, but trained professionals must remain responsible for calls. He cites rapid translation, realistic training, broader call-quality review and help with routine non-emergency calls—which account for roughly two-thirds of calls to 911 centers—as potential uses. He also calls for clear standards, continued investment in people and infrastructure, federal support for a nationwide system transition, and recognition of telecommunicators as protective service professionals.

AI agents, which can pursue goals and carry out multi-step tasks through software or other agents, raise questions about security and oversight. Companies use sandboxes to restrict access during testing, while the UK AI Security Institute has tested agents interacting with real people and systems under controlled conditions, deliberately granting internet access for cyber challenges. Researchers warn that agents could pose risks if they affect critical infrastructure or industrial processes; human approval can be difficult to position, and AI can assist cybercriminals and state-backed groups with reconnaissance, coding and phishing.

Key Takeaways

  • Google plans to make Guided Learning the default for K–12 search in January; students may lose access to Google Search if their districts do not permit it, amid concerns about unreliable AI answers.
  • A Norwegian committee recommends more supervised in-person university exams, phasing out short take-home exams, and strengthening lecturers’ AI skills.
  • OpenAI’s Codex promotion at TwitchCon sparked debate over AI and creator rights; Dan Clancy says Twitch’s opt-out setting expresses creators’ preferences about possible Amazon AI training, but Amazon’s use of Twitch content remains unclear.
  • Kukoon Rugs and Seating Matters use AI for work ranging from product design and supply chains to accounting and information gathering, while emphasizing human judgment and safeguards.
  • Carahsoft and more than 50 partners will present AI and high-performance computing solutions at SC26; Animashree Anandkumar is the keynote speaker, and Carahsoft will be at Booth #2006.
  • A study across 56 medical specialties found greater physical work in dermatology, orthopedics and ophthalmology, while pathology and diagnostic radiology ranked among the most exposed to AI substitution.
  • Google’s Chubby design emphasizes shared coordination, simplicity, leases, sequencers and metadata—ideas that may help developers build reliable AI infrastructure.
  • Chess research found AI games remained more strategically complex later in play, while human players tended to simplify; the study did not show that tension causes stronger play.
  • John Provenzano says AI can support 911 translation, training and quality review, but trained telecommunicators should remain responsible for calls.
  • AI-agent testing highlights risks involving access to real systems, critical infrastructure and cyber misuse, alongside challenges in deciding where human approval belongs.

Google’s AI Search Plan Raises Concerns in California Schools

Google plans to block K-12 students from using its search engine starting in January unless their school districts allow its AI-powered version. The company says Guided Learning will become the default and help students work through problems, but AI Mode and AI Overview will remain part of search. Parents and education advocates criticized the plan, while Google says it is responding to educators who want more control. Common Sense Media found in 180 tests that Google’s AI Mode completed homework assignments and sometimes gave fake or unreliable answers. California schools are debating limits on AI, and Los Angeles Unified has already restricted generative AI.

Universities Urged to Build Staff AI Skills and Rethink Exams

A Norwegian government committee presented recommendations on artificial intelligence in higher education to Minister of Research and Higher Education Eileen Fugelsnes. Professor Anders Malthe-Sørenssen, who chaired the committee, said AI is changing how students complete work and how teachers assess it. The committee recommends more supervised, in-person exams and phasing out short take-home exams. It also says lecturers need the skills to teach students to use AI appropriately, and universities should assess how AI affects subjects and professional practice. The report warns that AI may harm learning without suitable teaching methods and gives students nine recommendations, including avoiding shortcuts and thinking independently.

AI Debate Follows OpenAI’s TwitchCon Booth

OpenAI promoted its Codex coding tool with a hackathon at TwitchCon in San Diego, drawing criticism from some attendees concerned about AI’s effect on creative work. Some streamers said tools such as speech-to-text, captions and clipping could help with accessibility, while others opposed AI content scraping and wanted clear ways to opt out. Twitch CEO Dan Clancy said the opt-out setting is meant to let creators express their wishes if Amazon wants to use Twitch content for AI training, but he said he was not involved in Amazon’s plans. It remains unclear whether Amazon has used Twitch content to train its models. Twitch Chief Product Officer Mike Minton declined to discuss past use, citing ongoing litigation.

Northern Ireland Businesses Put AI to Work in Traditional Industries

Kukoon Rugs, a family-owned business in Newry, County Down, uses AI across departments including finance, marketing, supply chain and product design. Chief Executive Paul Vallely says the technology helps the company work faster and more accurately, while reducing boring tasks and freeing staff for research and creative work. Seating Matters, a manufacturer of specialist clinical seating near Limavady, also uses AI in its daily operations. Managing Director Martin Tierney says AI helps with back-office tasks such as accounting, finance and gathering information, though he considers it weak in customer service. He encourages staff to use AI sensibly and within safeguards.

Carahsoft and Partners to Show AI and HPC Solutions at SC26

Carahsoft and more than 50 partners will present high-performance computing, hardware, artificial intelligence and education solutions at SC26. The conference will feature technical sessions, workshops and discussions on AI, data infrastructure, quantum innovation and other areas. Its theme, “HPC Unites,” highlights AI’s growing role across scientific fields and the value of shared knowledge and technology in research. Caltech professor and Accelerated Understanding co-founder Animashree Anandkumar will give the keynote, “The Next Scaling Law after LLMS: AI that Understands the Physical World.” Visitors can meet Carahsoft at Booth #2006.

Study maps which medical specialties face the most AI exposure

Vivek Subbiah of NYU Langone Health shared research on how physical work may limit AI’s ability to replace tasks in different medical specialties. The study found that dermatology, with 68% physical work, orthopedic surgery at 53%, and ophthalmology at 45% were among the most shielded. Pathology at 99%, diagnostic radiology at 96%, and radiation oncology at 62% were among the most exposed. Across 56 specialties, 24% of work involved physical tasks, 21% was nonphysical and potentially substitutable, and 55% involved Evaluation and Management services.

Google’s Chubby design offers lessons for today’s AI systems

Google introduced its Chubby Lock Service in a 2006 paper by Mike Burrows. Chubby was built for reliable, low volume storage and coarse grained advisory locks, with a focus on simplicity rather than speed. Its key idea was to make coordination a shared service instead of requiring each application to build its own. The article explains how ideas such as leases, sequencers, and shared metadata may help developers build reliable AI infrastructure, including systems with distributed pipelines, vector databases, inference gateways, and GPU workers.

Chess study finds AI keeps strategic tension high

Researchers compared human chess games with games between Stockfish and Leela, measuring the number of possible interactions between opposing pieces. After the first 20 moves, human games had greater average tension, but human players later reduced it while AI games stayed more complex. The researchers say people may simplify positions to manage the many possibilities, while AI engines do not face the same limit. Tension also rose with higher human ratings, more thinking time, and deeper searches by Stockfish, though the study did not show that tension caused stronger play.

AI can help 911 teams while people stay in charge

John Provenzano, CEO of The 9-1-1 Association, says AI can make emergency communications faster and more reliable, but trained 911 professionals must remain responsible for calls. AI can translate languages quickly, support realistic training, review more calls for quality assurance, and help handle routine non-emergency calls, which make up roughly two-thirds of calls to 911 centers. Provenzano says Congress should recognize telecommunicators as protective service professionals and support funding for the nationwide transition to an updated system. He also calls for clear standards, responsible use, and continued investment in people and infrastructure.

AI agent tests raise questions about security and control

AI agents use large language models to pursue goals, handle multi-step tasks, and interact with software or other agents, sometimes without direct human involvement. Companies use sandboxes to limit agents’ access to files, networks, and other resources during tests. The article says the UK AI Security Institute found agents interacting with real people and systems during controlled challenges, but researchers had deliberately given them internet access to test their cyber abilities. AI agents could create risks for critical infrastructure if their actions affect machinery or industrial processes. The article also notes that human approval can be difficult to place, and that AI can help cybercriminals and state-backed groups with tasks such as reconnaissance, coding, and phishing.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI in education Google Guided Learning AI assessment AI and academic integrity AI in creative work creator rights AI training data business AI adoption high-performance computing AI infrastructure medical AI exposure distributed systems AI and chess emergency communications AI agents AI security cybersecurity human oversight

Comments

Loading...