Microsoft Says Agent 365 Will Manage Third Party and Local Agents

Several studies focus on the records and governance needed to make AI systems accountable. A proposed framework for clinical AI audits compares source records, information provided to the model, its output and subsequent clinical actions, drawing on HL7 FHIR Release 5 Provenance and AuditEvent resources and SMART frameworks. Its authors stress that an AI-generated explanation cannot substitute for objective records. Separately, Yan Li’s DRPG-AI links data, transformations, models, predictions and legal policies in a provenance graph. Tested on UCI Adult and COMPAS, it recorded a 0.8763 risk-detection F1 score, 0.9180 AUC and 0.9561 compliance coverage, outperforming 12 audit and fairness baselines.

Other reviews point to organizational and data barriers to AI adoption. An analysis of 70 ESG-finance studies published from 2017 to 2025 identifies inconsistent disclosures, incompatible metadata and proprietary data—not computing power—as obstacles to reliable AI; half the studies appeared in 2024 and 2025. A review of 75 supply-chain studies finds that predictive analytics, machine learning, digital twins and other AI tools can support resilience through agility and adaptability, provided organizations have digital readiness and strong governance. The review notes that lean supply chains can be exposed to disruptions including pandemics, geopolitical conflict and extreme weather.

Consumer-facing AI is also drawing scrutiny. In tests a week before Black Friday 2026, ChatGPT, Claude and Google’s Gemini gave unreliable results for 55-inch TV deals, including stale or dead links, inaccurate prices, unavailable stock or missing retailer links. A YouGov survey across 19 countries found the UK second-last for shoppers using AI, while 36% of respondents planning to shop around Black Friday said they would not use an AI tool. Amazon launched Alexa for Shopping in the UK in November, and OpenAI introduced ChatGPT Shopping Research shortly before the previous Black Friday.

In Raleigh, tenant Laura Mack says an AI agent named Olivia at Colony Townhome Apartments emailed her on Nov. 8 about paying nearly $4,000 in overdue rent by Nov. 30. Mack says she later received an eviction notice and was evicted; the emails did not disclose that they were AI-generated. North Carolina Sen. Jim Burgin says he plans to renew a proposed Bot Bill requiring disclosure when people interact with AI. The complex said it was gathering information. Separately, former FBI agent Tom Simon warns that criminals may use AI-created identities in home-title theft, which can prevent owners from selling or refinancing. He recommends monitoring credit reports; the article cites more than 100,000 reported cases in 2020, up 20% year over year.

Companies are weighing the practical costs and controls of AI. At a Dell Technologies event in New York, speakers discussed budgeting for token use, moving some workloads onto company systems and distributing inference across cloud, data-center and edge infrastructure. Comet Cloud offers dedicated GPU clusters, while Alpha Compute focuses on data sovereignty, confidential computing and managed infrastructure. At Microsoft Ignite on Nov. 17, 2026, Microsoft said Agent 365 will manage supported third-party and local agents, including integrations with Salesforce Agentforce, Amazon AgentCore and Google Vertex AI. Rollout is expected to finish in December; Microsoft says more than 50 million agents have been registered since May. Defender can find more than 60 local agents on managed Windows and macOS devices, while some discovery features remain in preview.

AI’s commercial promise and risks are also shaping markets and legal debates. Novo founder Michael Rangel says small businesses are only beginning to adopt AI at scale and predicts Anthropic’s IPO could spur an entrepreneurial gold rush. Yet the IPO did not calm markets: the Nasdaq fell 5.6% and the S&P 500 3.7%. AMD dropped 13%, Intel 8%, Broadcom 12% and Nvidia 14%, amid concerns about AI infrastructure costs and borrowing; regulators were examining Nvidia’s accounting practices. In Australia, Shine Lawyers is investigating a possible class action alleging OpenAI, Microsoft, Meta, Google and Anthropic used creative work without permission to train or run generative AI. California Attorney General Rob Bonta says the state is independently reviewing mergers, securities and AI regulation and will act on antitrust violations it finds.

Key Takeaways

  • A clinical AI audit proposal compares source records, model inputs and outputs, and later clinical actions; its authors say generated explanations do not replace objective records.
  • DRPG-AI achieved a 0.8763 F1 score, 0.9180 AUC and 0.9561 compliance coverage on UCI Adult and COMPAS tests, outperforming 12 audit and fairness baselines.
  • A review of 70 ESG-finance studies identifies inconsistent disclosures, incompatible metadata and proprietary data as major AI obstacles; half the studies were published in 2024–25.
  • A 75-study supply-chain review says AI supports resilience through agility and adaptability when backed by digital readiness and governance.
  • Tests of ChatGPT, Claude and Gemini found unreliable Black Friday 55-inch TV deal information; 36% of surveyed shoppers planning to shop around said they would not use AI.
  • After a Raleigh tenant said an apartment AI agent’s emails did not disclose its identity before an eviction, North Carolina Sen. Jim Burgin plans to renew a bill requiring AI-interaction disclosure.
  • Former FBI agent Tom Simon warns AI-generated identities may aid home-title theft and recommends monitoring credit reports; the article reports more than 100,000 cases in 2020.
  • Microsoft plans to expand Agent 365 to supported external and local agents, including Amazon AgentCore and Google Vertex AI integrations; it says over 50 million agents have been registered.
  • The Nasdaq fell 5.6% and the S&P 500 3.7% despite Anthropic’s IPO; AMD fell 13% and Intel 8% as investors weighed AI infrastructure costs and borrowing concerns.
  • Shine Lawyers is investigating possible Australian copyright claims involving OpenAI, Microsoft, Meta, Google and Anthropic; California AG Rob Bonta says the state is pursuing independent antitrust and AI reviews.

Researchers Outline a Framework for Auditing Clinical AI Decisions

Gerson Hiroshi Yoshinari Jr., Sandeco Macedo, and Luciano Magalhães Vitorino propose a framework for checking whether records can show how clinical AI decisions were made. They distinguish a correctly formatted or accurate result from one that can be traced through the system and reviewed. The framework compares source records, information sent to the AI, its output, and later clinical actions, while warning that AI-generated explanations do not replace objective records. It builds on existing tools, including HL7 FHIR Release 5 resources Provenance and AuditEvent, and SMART frameworks.

Review Finds Data Gaps Hold Back AI in ESG Finance

A review of 70 peer-reviewed studies published from 2017 to 2025 found that data problems, not computing power, are the main barriers to dependable AI in environmental, social, and governance finance. Researchers Miehleketo Mathebula, Abiodun Modupe, Vukosi Marivate, and Elda Du Toit identified unstandardized disclosures, incompatible metadata, and reliance on proprietary data as key limits. The team searched four databases using PRISMA guidelines and assessed the studies with a six-part quality framework. Two independent reviewers scored the research and reached a Cohen’s kappa agreement of 0.82. Half of the reviewed studies appeared in 2024 and 2025.

AI Can Strengthen Supply Chains When Organizations Build Agility and Adaptability

A review of 75 studies published from 2015 to 2025 found that AI can support supply chain resilience when organizations use it as a governed capability, not as a standalone tool. Adel Aazami, Amirhossein Hassani, Nadun Karunarathna, and Sebastian Kummer examined technologies including predictive analytics, machine learning, deep learning, digital twins, and robotic process automation. They found that AI contributes to resilience through two organizational paths, agility and adaptability, and depends on digital readiness and strong governance. The review notes that lean supply chains can be vulnerable to major disruptions such as the COVID-19 pandemic, geopolitical conflicts, and extreme climate events.

DRPG-AI Uses Provenance Graphs to Improve AI Compliance and Risk Detection

Researcher Yan Li developed DRPG-AI, a framework that uses machine learning records to support risk checks and regulatory compliance. In tests on the UCI Adult and COMPAS datasets, it achieved a risk-detection F1-score of 0.8763, an area under the receiver operating characteristic curve of 0.9180, and compliance coverage of 0.9561. It outperformed 12 audit and fairness baselines, including the strongest composite baseline, Themis coupled with W3C PROV. The framework connects data, transformations, models, predictions, and legal policies in a directed provenance graph to help trace operational risks and compliance issues.

AI Shopping Tools Struggle to Find Reliable Black Friday TV Deals

The article tested free versions of four AI tools a week before Black Friday 2026 by asking them to find a 55-inch TV deal. ChatGPT, Claude, and Gemini gave mixed results, including unavailable products, dead links, incorrect prices, or missing store links. ChatGPT recommended a Hisense TV and TCL models, but some offers were unavailable or over budget; Claude gave limited details and links to dead pages; Gemini listed prices that did not always match retailers. The article says a YouGov survey across 19 countries found the UK ranked second-last for shoppers using AI, and 36% of respondents planning to shop around Black Friday did not plan to use an AI tool. It also notes that Amazon launched Alexa for Shopping in the UK in November and OpenAI introduced ChatGPT Shopping Research days before the previous Black Friday.

AI tools struggle to find reliable Black Friday TV deals

Reporters tested four AI tools a week before Black Friday to see if they could find a deal on a 55-inch TV. ChatGPT suggested TVs that were out of stock, not listed online or over budget, while some links led to unavailable products. Claude gave little detail and linked to dead pages or old Black Friday 2025 lists. Gemini compared four models, but did not provide store links, and some prices or stock information were unreliable. The article says shoppers are beginning to consider AI for shopping, as companies including Amazon, OpenAI and Google add shopping features.

Raleigh eviction spurs renewed push for AI consumer protections

Laura Mack says an AI agent named Olivia at Colony Townhome Apartments emailed her on Nov. 8 that she could pay nearly $4,000 in overdue rent by Nov. 30. Mack says she later received an eviction notice and was evicted on Thursday. The emails did not say they were written by AI, and Mack thought she was speaking with someone from the leasing office. North Carolina state Sen. Jim Burgin says the case shows the risks of AI in business decisions and plans to renew his proposed Bot Bill, which would require users to be told when they interact with AI. The apartment complex said it is gathering information and will update WRAL the following week.

Former FBI agent warns AI may be used in home title theft

Former FBI special agent Tom Simon says criminals are using AI to create fake identities and steal home titles. He warns that this crime can stop homeowners from selling their homes or refinancing their mortgages. Simon advises homeowners to stay alert and monitor their credit reports. He also recommends using a title insurance company that uses AI to detect possible problems. The article says more than 100,000 home title thefts were reported in 2020, up 20% from the year before.

Dell event highlights three steps for AI operational readiness

A Dell Technologies event in New York explored what companies need to prepare for AI, beyond choosing servers, storage and computing capacity. As AI use grows, finance teams are setting budgets for token use and weighing whether some workloads can run more cheaply on company systems. Companies are also spreading AI workloads across cloud services, data centers and edge devices to manage the cost of inference. Speakers also discussed how smaller businesses can get access to GPU computing, with Comet Cloud offering dedicated clusters and Alpha Compute focusing on data sovereignty, confidential computing and managed infrastructure services.

Anthropic IPO fails to stop tech stocks from plunging

The tech-heavy Nasdaq fell 5.6%, wiping out its gains for the year, while the S&P 500 dropped 3.7%. Anthropic’s major IPO did not ease investor worries about AI, including the high cost of infrastructure and possible limits on borrowing as interest rates rise. Chip stocks fell sharply, with AMD down 13%, Intel down 8%, Broadcom down 12% and Nvidia down 14%; regulators said they were examining Nvidia’s accounting practices. Investors have also shifted money from technology stocks to safer areas, while some analysts say tech shares may be undervalued because projected earnings remain strong. Interest in Anthropic shares among retail investors may become clearer when options begin trading on Monday.

Australian artists consider class action over AI copyright use

Shine Lawyers is investigating whether Australian visual artists, authors and musicians could bring a class action against technology companies. The firm is examining claims that OpenAI, Microsoft, Meta, Google and Anthropic copied music, lyrics, sound recordings and visual art without permission to train and run generative AI systems. Shine class actions head Craig Allsopp said artists and musicians deserve to know whether companies used their work without consent.

Microsoft expands Agent 365 controls to outside and local AI agents

At its November 17, 2026, Ignite event, Microsoft announced that Agent 365 will manage supported third-party AI agents and agents running on employees’ devices. Connected agents can receive Microsoft Entra Agent IDs, and administrators can view details such as their owners and tools and apply supported access controls. Integrations include Salesforce Agentforce, Amazon AgentCore and Google Vertex AI, with rollout expected to finish in December. Microsoft Defender can find more than 60 local agents on managed Windows and macOS devices, while network discovery and registry entries for unapproved agents remain in preview. Microsoft says organizations have registered more than 50 million agents since Agent 365 launched in May, and premium features for outside agents require Agent 365 or ME7.

Novo founder sees small business AI boom after Anthropic IPO

Michael Rangel, founder of small business financial technology platform Novo, said small businesses are only beginning to adopt AI at a large scale. He predicts that Anthropic’s IPO will help spark an entrepreneurial gold rush. The article describes his view of the opportunity for small businesses as AI adoption grows.

California attorney general plans independent antitrust and AI reviews

California Attorney General Rob Bonta told POLITICO that the state is examining mergers, corporate securities and AI regulation on its own. He said California would investigate whether companies follow antitrust laws and act if it finds violations. Bonta’s doubts about federal antitrust regulators grew after the Justice Department settled its case against Live Nation Entertainment in March. The lawsuit, brought with 40 states and the District of Columbia, sought to break up the company’s control of live events and require it to sell Ticketmaster.

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 accountability AI auditing data provenance clinical AI HL7 FHIR AI governance AI adoption ESG finance supply chain resilience generative AI AI shopping assistants AI disclosure AI-enabled fraud AI infrastructure cloud computing edge computing AI agents Microsoft Agent 365 AI market volatility AI copyright

Comments

Loading...