Google Anthropic and OpenAI lead AI news

AI is spreading into public services and workplaces, while researchers and advocates press for clearer safeguards. An eight-year project led by Professor Zhong-Ren Peng distinguishes Urban AI, which studies city systems, from Urban Planning AI, which can influence zoning, land development and public resources. A 2024 review of 881 papers described four levels of AI use in planning, from AI-assisted to AI-autonomized; a 2026 paper in the Journal of the American Planning Association advances proposals for governing its use. In London, where Google DeepMind and Anthropic are based, the London AI Hub says as many as 50 organizations focus on AI safety. The non-profit London Initiative for Safe AI plans to move to a larger headquarters in King’s Cross.

Questions of accountability also arise as AI agents gain authority to act. Harvard Kennedy School fellow Shlomit Wagman says payment systems lack clear rules for what agents may decide and who pays when they misjudge a user’s instructions. For example, an agent could stay within a $5,000 vacation budget yet make purchases that do not meet the user’s wishes. Wagman calls for verifiable mandates, clear liability and enforceable duties to put users first. Separately, a cybersecurity report finds 96% of surveyed organizations use agents that can act for employees, and 52% use them widely. About four in five give agents access to internal systems; 77% say an agent or tool acted on malicious, misleading or unauthorized content in the past year. Only 6% name security teams as the main approvers, compared with 78% that assign approval to IT.

At Los Alamos National Laboratory, a year into its partnership with OpenAI, use of the secure Venado-GPT service has surged. Staff at Los Alamos, Lawrence Livermore and Sandia can access three experimental models on the Venado supercomputer. From mid-February to late August 2026, weekly token use rose 20-fold to more than 3 billion, and successful weekly requests grew from about 1,200 to over 41,200. Uses include certification, translating simulation models, digital-twin workflows and network maintenance. The partners plan improvements to file reading, API reliability and complex coding support. Meanwhile, more than 550 Google employees petitioned for schools to be able to use Search without AI; parents and educators made the same request at protests outside Google offices in six cities.

Research and tools are also targeting agriculture. A review of 191 open-source farm tools by the EU-funded Horizon Europe OpenAgri project found that tools directly automating farm work tend to use rule-based algorithms and offline edge computing, rather than cloud deep learning. The researchers cite scattered information, limited digital skills and rural connectivity gaps as barriers to wider adoption. Penn State researchers, in findings published in Scientific Reports, developed PSUMamba, which uses nondestructive imaging to detect crop stress from drought, disease, nutrient shortages and other causes, potentially with less computing power than some existing methods.

Efforts to broaden participation in AI governance include a partnership between Africa Wiki Women and GIRAI. Their plans include community learning sessions, Wikimedia content campaigns, podcasts and participation in conferences and advocacy events. In July 2026, Africa Wiki Women members joined the online unveiling in Geneva of the second Global Index on Responsible AI, which assessed 135 countries and jurisdictions on areas including inclusion, ethics, trust and safety, and public services. In another field, a November 20, 2026 Business Insider article says debate is intensifying over whether mathematicians or AI will prove the Riemann Hypothesis, but gives no details about the competing attempts.

AI is also entering real estate photography: a photographer testing Fotello said it naturally neutralized a bright wall color, blended in a custom sky and handled other detailed edits. The article raises, but does not resolve, whether such tools will replace photographers or help them work more efficiently.

Key Takeaways

  • A 2026 planning paper builds on research reviewing 881 studies and proposes governance for AI that can affect zoning, development and public resources.
  • London’s AI safety network includes up to 50 organizations; London Initiative for Safe AI plans to expand its King’s Cross headquarters. Google DeepMind and Anthropic are based in the city.
  • Harvard Kennedy School fellow Shlomit Wagman calls for verifiable user mandates, clear liability and enforceable duties for AI payment agents.
  • A cybersecurity report says 96% of surveyed organizations use AI agents, 77% experienced agents or tools acting on malicious or unauthorized instructions, and only 6% make security teams the main approvers.
  • Venado-GPT’s weekly token use grew 20-fold to more than 3 billion between mid-February and late August 2026; weekly successful requests rose from about 1,200 to more than 41,200.
  • More than 550 Google employees petitioned for schools to have a Search option without AI, echoing requests from parents and educators at protests in six cities.
  • A review of 191 open-source agriculture tools found direct farm automation relies mainly on rule-based algorithms and offline edge computing; connectivity and digital-skills gaps remain obstacles.
  • Penn State’s PSUMamba detects crop stress through nondestructive imaging and may use less computing power than some existing AI methods.
  • Africa Wiki Women and GIRAI plan community education and advocacy on responsible AI; the second Global Index on Responsible AI assessed 135 countries and jurisdictions.
  • Fotello produced realistic real-estate photo edits in a photographer’s test; a Business Insider article reported intensifying debate over AI and the Riemann Hypothesis without detailing the competing attempts.

Framework sets rules for AI in city planning

Professor Zhong-Ren Peng’s eight-year research project examines how cities can use artificial intelligence while protecting public accountability, democratic decision-making and fairness. It distinguishes Urban AI, which studies city systems, from Urban Planning AI, which can affect public resources, zoning and land development. A 2024 study reviewed 881 research papers and described four levels of AI use in planning, from AI-assisted to AI-autonomized. A 2026 paper in the Journal of the American Planning Association advanced the work toward rules for governing AI in planning.

Open-source farm tools often use offline rules to automate work

Researchers in the European Union-funded Horizon Europe OpenAgri project reviewed 191 open-source digital tools for agriculture. The study, published in Smart Agricultural Technology, found that tools with the most direct farm automation mainly use rule-based algorithms and offline edge computing, rather than cloud-based deep learning. The catalog included software and open hardware, and researchers gathered examples from sources such as GitHub, FarmHack and the Open-Source Hardware Association directory. The article also describes barriers to wider use, including scattered information, limited digital skills and rural connectivity gaps.

AI payment agents need clear rules for mistakes and liability

Shlomit Wagman, a fellow at Harvard Kennedy School’s Mossavar-Rahmani Center for Business and Government and founder of Integria AI, argues that payment systems do not yet address what AI agents may decide or who bears the cost when they make poor choices. An agent could follow a $5,000 vacation budget and complete authorized purchases that still fail to meet the user’s instructions. Wagman says systems need verifiable mandates that capture users’ goals, a clear plan for assigning liability, and enforceable duties to put users’ interests first. She notes that existing payment protections do not clearly settle responsibility for mistakes in an agent’s judgment.

Penn State AI model detects crop stress with less computing

Researchers at Penn State developed an AI model called PSUMamba to detect stress in crops using nondestructive imaging. The model can help monitor crop health over time and detect stress caused by drought, disease, nutrient shortages and other causes. The researchers say it may provide accurate results while using less computing power than some existing AI methods. They published their findings in Scientific Reports.

Mathematicians and AI enter a race to prove the Riemann Hypothesis

A Business Insider article describes a growing human-versus-AI contest over proving the Riemann Hypothesis. The story was published on November 20, 2026, and written by Truman Dickerson, a weekend news fellow based in New York City. The supplied article text says the debate in the math community is intensifying but provides no further details about the competing attempts.

London builds a growing network focused on safer AI

London has become a major AI hub, with Google DeepMind and Anthropic based in the city. Researchers, policymakers and campaigners are working to make AI systems benefit society and avoid harm. The London AI Hub says up to 50 local organisations now focus on AI safety. One of them, the non-profit London Initiative for Safe AI, plans to move to a larger headquarters in King’s Cross.

Los Alamos and OpenAI report rising use of AI models

One year after starting their partnership, Los Alamos National Laboratory and OpenAI report growing use of AI models on Venado, Los Alamos’ supercomputer. Through Venado-GPT, staff at Los Alamos, Lawrence Livermore and Sandia can use three experimental models in a secure computing environment. From mid-February to late August 2026, weekly token use grew 20-fold to more than 3 billion, while weekly successful requests rose from about 1,200 to more than 41,200. Researchers have used the service for certification work, simulation-model translation, digital-twin workflows and network maintenance. The partners plan to improve file reading, API dependability and support for complex coding workflows.

African Wiki Women join efforts to shape responsible AI

Africa Wiki Women and GIRAI are working together to help African communities, especially women, take part in discussions about responsible AI. The partnership connects African open knowledge communities with GIRAI’s work assessing responsible AI policies and practices across countries. Plans include community learning sessions, Wikimedia content campaigns, podcasts, and participation in conferences and advocacy events. In July 2026, AWW members joined the online unveiling of the second Global Index on Responsible AI in Geneva. The Index assessed 135 countries and jurisdictions across areas including inclusion, ethics, trust and safety, and public services.

Fotello AI edits real estate photos with realistic results

A photographer tested Fotello, an AI-powered photo editing tool for real estate photography. The tool neutralized a bright wall color, blended a custom sky into an image and made other detailed edits. The photographer said the results looked natural and appeared to have been done by a person. The article raises questions about whether tools like Fotello will replace photographers or help them work more efficiently, but says those questions will be explored in future articles.

AI agents bring new cybersecurity risks to businesses

A cybersecurity report says 96% of surveyed organizations use AI agents that can act for employees, and 52% use them widely. About four in five give these agents access to internal systems, including help desks, cloud storage, email and finance tools. In the past year, 77% of respondents said an AI agent or tool acted on malicious, misleading or unauthorized instructions or content. Common incidents involved tools handling malicious messages and attackers using AI in phishing attempts. Only 6% of organizations said security teams were the main approvers of AI agents, while 78% assigned that role to IT teams.

Google workers push for schools to use Search without AI

More than 550 Google employees signed a petition asking the company to let schools use Search without AI. Parents and educators made the same request during protests outside Google offices in six cities last week. The article is dated November 21, 2026.

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.

Urban AI AI governance AI in urban planning AI safety AI agents Agentic AI security AI accountability AI liability Venado-GPT OpenAI AI in scientific research AI in agriculture Edge computing PSUMamba Responsible AI AI inclusion AI in education AI search AI in real estate photography AI and mathematics

Comments

Loading...