Microsoft 365 outage hits millions as Nvidia chips appear in AI models

Microsoft says a global Microsoft 365 outage that began late November 12 stemmed from a problem in an earlier network-capacity upgrade that surfaced after a later network change—not a cyberattack, malicious activity or rogue AI. Outlook, Teams and Copilot were among the affected services, disrupting millions of users, including banks, government departments and courts in the United States and Japan. Microsoft said its first repair changes were not working as expected on November 13 and that it was pursuing other options; the disruption continued for more than 24 hours.

Two AI-security deals are advancing. Startup Tensor plans to acquire about 90% of Nasdaq-listed Rain Enhancement Technologies Holdco (RAIN) through a cash-and-stock reverse merger, subject to approvals; RAIN shareholders will retain the balance and receive cash. The deal would make RAIN the first Nasdaq-listed company acquired by a private entity through a reverse merger. Separately, Vanta has agreed to acquire Pillar Security and add its AI-security technology to Vanta’s cloud security and compliance platform. That deal also awaits regulatory approval.

Investors in AI infrastructure face possible policy and power constraints in 2026. A political shift after the midterm election could bring more scrutiny and restrictions, raising costs for data centers and other projects. Key questions include whether projects can secure electricity, how rules affect siting, and whether approved or underway projects are protected. Large technology companies may be better positioned because of their resources, existing facilities, utility relationships and secured power capacity.

NVIDIA’s 2026 report says 91% of retail and consumer-goods companies are using or evaluating AI, and 58% have put it to active use. Among companies using AI, 67% create marketing or advertising content with it; other common applications are product recommendations (58%), ad placement (54%) and adaptive advertising (52%). Of companies using AI agents, 48% apply them to personalized marketing and advertising.

Deloitte identifies active ETFs, broader private-market access, changing talent needs and firm-wide AI platforms as forces shaping investment management in 2027. Active ETFs drew 26% of U.S. ETF net inflows in 2024, up from 1% in 2014, while assets rose from $502 billion to $843 billion. The report says alternative-fund assets could grow by more than 50% annually to as much as $4.1 trillion by decade’s end. U.S. investment-management job postings mentioning AI rose nearly 25% between 2022 and the first half of 2025; Deloitte’s Vik Bhat points to investment in data, training and risk management as potential advantages.

A study by former OpenAI researcher Will DePue reports Nvidia Blackwell hardware fingerprints in six public models from DeepSeek, Moonshot AI, MiniMax and Ant Group. After comparing more than 6 trillion chip-and-random-seed combinations and testing against 24 models with known hardware, DePue estimated Kimi K3 and DeepSeek-V4-Pro may each have used 2,000–3,000 Blackwell GPUs, potentially worth more than $500 million. The findings do not establish that the chips entered China or that export rules were violated; their source and location remain unclear.

South Pasadena Unified School District adopted a generative-AI policy and a three-year strategic plan on November 10, 2026. Students may use AI on school devices only when a teacher assigns it, must disclose its use, and remain subject to existing plagiarism rules. The board also authorized up to $46 million in Measure SP bonds, approved an independent citizens’ oversight committee and 2026–27 agreements covering raises for teachers and staff. Dr. Michele Kipke delivered her final board report after nearly 20 years of service.

Cloudflare says its AI testing system can run penetration tests through agents that map assets and take sequences of actions to find and exploit weaknesses. The company recommends layered defenses including web application firewalls, rate limits, anomaly detection, bot protection, identity checks, validated data formats and threat intelligence. The article notes that complex attacks can combine stolen customer credentials with movement through a network; system maintainers say the tool is for defensive use.

Two other reports examine the human and oversight sides of AI. Epidemic Sound’s Gracie Schram says high-earning creators face pressure to meet rising production and distribution expectations, including pressure to use AI. Researchers’ ReA-Bench evaluates whether auditors can spot risks in autonomous-agent actions, addressing missed risks and mistaken flags on safe behavior. Its six challenge tiers cover safety knowledge, contradictions, long-context evidence, source checks, indirect steering and missing information; it uses realistic, mostly production-based trajectories, including difficult benign cases and paired versions to test the effects of context.

Key Takeaways

  • Microsoft traced the outage affecting Outlook, Teams and Copilot to a network-capacity issue, not a cyberattack or rogue AI; millions of users were affected and initial repair changes failed to work as expected.
  • Tensor plans to acquire about 90% of Nasdaq-listed RAIN in a cash-and-stock reverse merger; Vanta separately agreed to acquire AI-security platform Pillar Security.
  • Post-election policy changes could increase AI infrastructure costs, with electricity access, project approvals and data-center siting among the key issues.
  • NVIDIA reports 91% of retail and consumer-goods companies are using or evaluating AI; 67% of AI users apply it to marketing or advertising content.
  • Deloitte says active ETFs grew from $502 billion in assets in 2014 to $843 billion, and received 26% of U.S. ETF net inflows in 2024; alternative funds could reach $4.1 trillion by decade’s end.
  • Former OpenAI researcher Will DePue reports possible Blackwell fingerprints in models including DeepSeek’s; estimated GPU use for DeepSeek-V4-Pro and Kimi K3 could exceed $500 million in value, but the study does not establish chip location or export-rule violations.
  • South Pasadena Unified’s AI policy requires teacher assignment for use on student devices and disclosure by students; the board also authorized up to $46 million in Measure SP bonds.
  • Cloudflare describes AI agents conducting end-to-end penetration tests and recommends layered defenses; its system maintainers say it is for defensive use.
  • ReA-Bench tests auditors’ ability to identify risks in autonomous-agent behavior, including missed risks and false alarms on benign actions.
  • Epidemic Sound says high-earning creators face pressure to meet production and distribution expectations, including expectations to use AI.

Microsoft rules out cyberattack and AI as cause of global outage

An outage affecting Outlook, Teams and Microsoft Copilot began Thursday and continued for more than 24 hours as of Friday morning, affecting millions of customers worldwide. Microsoft said the disruption was caused by a problem during an earlier network capacity upgrade that appeared after a recent network change. The company said it found no signs of malicious activity or a cyberattack.

Microsoft rejects cyberattack and rogue AI theories behind outage

A Microsoft 365 outage that began late November 12 affected millions of people worldwide who use Outlook, Teams and other services. Banks, government departments and courts in countries including the United States and Japan were among those affected. Microsoft said there was no indication the outage resulted from a cyberattack or malicious activity, despite online theories blaming a cyberattack or a rogue AI agent. On November 13, the company also said its first repair changes were not working as expected and it was looking at other options.

AI security startup Tensor to acquire 90% of Nasdaq-listed RAIN

AI security startup Tensor is set to acquire about 90% of Rain Enhancement Technologies Holdco, or RAIN, in a reverse merger. The deal would make RAIN the first Nasdaq-listed company acquired by a private entity through a reverse merger. Tensor will pay with a combination of cash and stock, while RAIN’s existing shareholders will keep the remaining shares and receive cash. The deal is expected to close in the coming weeks, subject to regulatory approvals and other usual conditions.

Vanta agrees to acquire AI security platform Pillar Security

Vanta has signed an agreement to acquire Pillar Security, an AI security platform. Vanta says Pillar’s technology will be added to its cloud-based security and compliance platform to help businesses protect against emerging threats. The acquisition is subject to regulatory approval and is expected to close in the coming months. The companies describe the deal as a way to expand Vanta’s security and compliance offerings.

Election could bring tougher rules and higher costs for AI investment

AI investment in 2026 is driving major spending on data centers, semiconductors and other infrastructure. The article says a shift in the political environment after the midterm election could bring more scrutiny and restrictions, raising costs for new projects. Investors should watch whether projects can secure enough electricity, whether approved or underway projects are protected, how rules affect where data centers are built, and whether higher costs slow spending. Large technology companies may be better placed because they have existing data centers, utility relationships, financial resources and secured electrical capacity.

AI Creates Marketing Content for 67% of Retail Companies Using It

A 2026 report by NVIDIA says 67% of retail and consumer goods companies using AI apply it to create marketing and advertising content. Other common uses include product recommendations at 58%, ad placement at 54%, and adaptive advertising at 52%. Among companies using AI agents, 48% use them for personalized marketing and advertising. Overall, 91% of companies in the sector are using or evaluating AI, and 58% have put it into active use.

AI and Active ETFs Reshape Investment Management for 2027

A Deloitte Center for Financial Services report identifies four forces shaping investment management in 2027: active ETFs, wider access to private markets, changing talent needs, and AI platforms used across entire firms. Active ETFs received 26% of US ETF net inflows in 2024, up from 1% in 2014, and their assets grew from $502 billion to $843 billion. The report says alternative funds could grow by more than 50% a year to reach as much as $4.1 trillion by the end of the decade. US investment management job postings mentioning AI rose nearly 25% from 2022 to the first half of 2025, while Deloitte’s Vik Bhat said firms could gain an edge by investing in data, workforce training, and risk management.

Study Links AI Model Weights to Nvidia Blackwell Chips

Former OpenAI researcher Will DePue says his study found hardware fingerprints linked to Nvidia’s Blackwell architecture in six publicly released AI models from DeepSeek, Moonshot AI, MiniMax, and Ant Group. He compared more than 6 trillion chip and random seed combinations and tested the method against 24 models with known hardware. DePue estimated that Kimi K3 and DeepSeek-V4-Pro may each have used 2,000 to 3,000 Blackwell GPUs, with the possible GPU value exceeding $500 million. The findings do not prove the chips entered China or that export rules were broken, and their location and source remain unclear.

South Pasadena Schools Adopt AI Policy and Approve $46 Million Bond

The South Pasadena Unified School District Board of Education adopted a generative AI policy and approved a new three-year Strategic Plan at its November 10, 2026 meeting. The policy blocks AI use on student devices unless a teacher assigns it, requires students to disclose its use, and keeps existing plagiarism rules in place. The board also authorized the sale of up to $46 million in Measure SP bonds and approved an independent citizens’ oversight committee. It approved 2026–2027 agreements covering raises for teachers and staff, while Dr. Michele Kipke gave her final board report after nearly 20 years of service.

Cloudflare Says AI Agents Can Run Full Penetration Tests

Cloudflare says its AI testing system can handle the full process of observing, planning, deciding, and acting without relying on human skill as the main limit. The company describes parallel agents mapping assets and using a sequence of actions to find and exploit weaknesses. Cloudflare recommends a security approach that includes web application firewalls, rate limits, anomaly detection, bot defense, identity checks, validated data formats, and threat intelligence. The article also says some complex attacks can combine stolen customer credentials with movement through a network, while the system’s maintainers say it is for defensive use only.

High earning creators face growing pressure to use AI

Epidemic Sound says creators who earn more face pressure to keep up with expectations for production and distribution. Gracie Schram of Epidemic Sound says this pressure includes using AI. The article focuses on the challenges high earning creators face as they work to meet these expectations.

ReA-Bench tests how well auditors spot risks in AI agents

Researchers introduced ReA-Bench, a benchmark for checking whether auditors can identify risks in the actions of autonomous agents. Their review of production and red team trajectories found that some risks are often missed, while overall recall does not show which risks are overlooked or whether auditors flag safe behavior by mistake. ReA-Bench groups challenges into six tiers, including safety knowledge, contradictions, long context evidence, source checks, indirect steering, and missing information. It includes difficult benign cases and paired trajectory versions to test whether shorter or more complete information helps auditors identify risks. Most samples come from real production trajectories, with scenarios selected and adapted to keep tasks and risks realistic.

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.

Microsoft 365 outage network capacity Outlook Microsoft Teams Microsoft Copilot AI security reverse merger Tensor RAIN Vanta Pillar Security AI infrastructure data centers AI policy AI in retail AI marketing active ETFs investment management NVIDIA Blackwell AI hardware exports

Comments

Loading...