Security
- Threat actors use agentic AI tooling to find and abuse exposed secrets
Microsoft has seen attackers use the ARTEX agentic framework to scan public GitHub repos for credentials and API keys, then act on them, in one case going from a GitHub scan to exfiltrating Azure Key Vault data in five minutes. Since 28 October it has also seen service principal sign-ins from 199.19.77[.]84 using stolen multi-tenant app credentials to reach downstream tenants. Hunt service principal and non-interactive sign-ins for the listed IPs in every tenant you manage, and rotate any secret that has ever touched a repo.
https://www.microsoft.com/en-us/security/blog/2026/11/05/threat-actors-use-agentic-ai-tooling-to-find-and-abuse-exposed-secrets/ - Inside a malicious Claude Code-branded extension and its persistent SSH backdoor
A tampered copy of the Claude Code extension (versions 2.1.312 and 2.1.313) sat in the Cursor extension marketplace for about 11 hours on 29 October, adding an attacker SSH key, enabling SSH, opening a reverse tunnel and setting up persistence. The Visual Studio Marketplace and npm packages weren’t affected. If your team or clients use Cursor, check for those versions and remember the keys and persistence can be left behind even when the tunnel failed.
https://www.microsoft.com/en-us/security/blog/2026/10/30/inside-malicious-claude-code-branded-cursor-extension-persistent-ssh-backdoor/ - Email threat landscape: Q3 2026 trends and insights
Microsoft detected about 7.3 billion email phishing threats in Q3, with September up 41% on August, and CAPTCHA-gated phishing coming back via SVG attachments. The standout finding is that around 129 million spoofed messages were delivered because org or user exceptions overrode protection even after they failed authentication. Good reminder to go and audit those “just allow this sender” exceptions in every tenant.
https://www.microsoft.com/en-us/security/blog/2026/11/05/email-threat-landscape-q3-2026-trends-and-insights/ - What exploited vulnerabilities have in common: A data-driven analysis of real-world attacks
Of roughly 383,000 CVEs disclosed since 1999, only about 2,000 (around 0.5%) have been exploited in the wild, and attackers overwhelmingly go after remote code execution, privilege escalation and authentication bypass. Patching is slow while exploitation is fast, and AI will only shorten that window. Worth a look for prioritising patching by impact and exposure rather than just CVSS score.
https://www.microsoft.com/en-us/security/blog/2026/11/02/exploited-vulnerabilities-common-data-driven-analysis-real-world-attacks/ - Fake-download campaign uses trojanized installers to deliver monetization payloads
People searching for cracked software, free games and activators are getting trojanised installers that silently turn their machines into proxy nodes, distributed-computing workers or hosts for bundled “offers”. Nothing groundbreaking, but it’s exactly the stuff that turns up on staff laptops. Turn on potentially unwanted application protection in Defender and block untrusted software sources.
https://www.microsoft.com/en-us/security/blog/2026/11/03/fake-download-campaign-uses-trojanized-installers-deliver-monetization-payloads/ - What’s new in Microsoft Security: November 2026
This month’s Intune and Defender roundup is heavy on agent security: Dynamic Permissions that scope an agent’s authority to its task, Dynamic Agent Risk and Dynamic Conditional Access, plus Microsoft Execution Containers (MXC), now generally available, to contain what agents can touch at the OS level. There are also endpoint elevation and Exposure Management updates. Still early for most SMBs, but this is the toolkit you’ll need once clients start running agents.
https://www.microsoft.com/en-us/security/blog/2026/11/05/whats-new-in-microsoft-security-november-2026/
Microsoft 365 & Windows
- Introducing SharePoint tools in Work IQ
More than 60 native SharePoint operations are now in public preview through the Work IQ MCP server, so Copilot, Copilot Studio and other agents can find, create, organise and share SharePoint content directly. Microsoft says task pass rates went from 70.1% to 84.6% while using 16% fewer tokens. Worth a look if you’re building agents for clients, but remember they’ll be acting on SharePoint with whatever permissions you give them, so tidy up oversharing first.
https://techcommunity.microsoft.com/blog/SPBlog/introducing-sharepoint-tools-in-work-iq/4569917 - How Outlook Search is changing: ranking, resilience, and the feedback loop
Outlook now shows up to seven relevance-ranked results before falling back to date order, across classic Outlook (Current and Monthly Enterprise Channel), new Outlook, the web and mobile, with Mac following in November. Classic Outlook also leans on the local Windows index when the network is flaky. Good to know when users ring the help desk saying “search is broken” and it’s really just ranking differently.
https://techcommunity.microsoft.com/blog/outlook/how-outlook-search-is-changing-ranking-resilience-and-the-feedback-loop/4568760
Cloud & AI
- Gemini 4 Carbon: our next frontier model
Google’s new frontier model leads on cybersecurity benchmarks for finding, exploiting and patching vulnerabilities, and is built for agentic tasks that run for days. Access is starting with a subset of cyber defenders and trusted partners, with wider release later at $2 per million input tokens and $10 per million output tokens. Still early, but models this good at exploitation are exactly why patch windows keep shrinking.
https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/ - Humanist AI in practice: A public consultation on our Code of Conduct for MAI Models
Microsoft AI has published a first draft of the code of conduct that governs how its MAI models are trained and how they should behave once deployed, and is asking for public feedback. It’s built around Mustafa Suleyman’s idea of AI that is “subordinate, aligned, and contained”. Nothing groundbreaking for day-to-day work, but it’s useful context when clients ask how Microsoft’s own models are governed.
https://microsoft.ai/news/mai-code-of-conduct/ - Fair use benefits everyone
OpenAI has rounded up the amicus briefs filed in support of fair use in the New York Times lawsuit, from library groups, the EFF and law professors through to the US Government. The outcome will shape whether AI training on copyrighted content stays lawful. Still early, and it’s one side’s argument, but it’s a case worth following because it affects the tools every client is adopting.
https://openai.com/index/fair-use-benefits-everyone/ - Giving every team a data scientist with Grok Bot
SpaceXSI’s data science team built Data Bot, a shared Team Bot that it teaches through skills covering metrics, experiments and query logic, so anyone can query the warehouse and get answers the way a data scientist would. Experts improve the skills once and everyone benefits. Worth a look as a model for MSPs: package your team’s know-how into a shared bot instead of being the bottleneck.
https://x.ai/news/data-bot - Elon Musk throttles Grok Bot limits due to ‘unsustainably high’ costs
Over Halloween weekend some Grok Bot users found their weekly usage limits suddenly maxed out, and Musk confirmed it was deliberate because the cost of serving it was “unsustainably high relative to revenue”. No word yet on what the limits will be going forward. Good reminder that AI pricing and quotas are still moving targets, so don’t build client services that assume today’s limits.
https://techcrunch.com/2026/11/02/elon-musk-throttles-grok-bot-limits-due-to-unsustainably-high-costs/ - How AI agents run insurance’s back office
Harper describes always-on ambient agents that wake on events, work on their own cloud computer and drive carrier websites that have no API, the way a broker would. Since September an agent has helped bring more than 200 small businesses current with their premium-finance lender, with a person still signing off before money moves. Swap “carrier portal” for “vendor portal” and you can see how much MSP back-office work this could take on.
https://www.harperinsure.com/engineering/ambient-agents/
Other
- What Happens to MSPs When Grok Bot Can Do the Work?
My latest post takes Harper’s account of running its back office on ambient agents and asks what that means for MSPs when ticket triage, remote repair, tenant checks and reporting can be packaged up and shared. If the instructions become a template anyone can download, the value has to sit somewhere else. I look at what’s left for the MSP, the vendor and the customer.
https://blog.ciaops.com/2026/11/06/what-happens-to-msps-when-grok-bot-can-do-the-work/ - How Grok Bot Can Resolve Help Desk Issues Through Remote Desktop
My latest post walks through having Grok Bot open a remote-support tool like TeamViewer or AnyDesk on its own computer, connect to an approved attended session and fix a user’s Outlook problem itself, then retest and write up the ticket. I also cover the guardrails: identity and consent checks, and treating inbound email as untrusted. It’s a workflow you configure and test, not something that happens just because the bot has an inbox.
https://blog.ciaops.com/2026/11/06/how-grok-bot-can-resolve-help-desk-issues-through-remote-desktop/ - One Prompt, Three Tenants: The Benefits of Grok Bot Group Chat
My latest post shows how I put three tenant bots into one Grok Bot group chat and ran Microsoft’s exposed-secrets hunting guidance across all of them with a single prompt. Each bot checked its own tenant and the answers came back side by side. I also explain why an invalid or expired secret error on its own doesn’t prove compromise.
https://blog.ciaops.com/2026/11/06/one-prompt-three-tenants-the-benefits-of-grok-bot-group-chat/
After hours
SpaceX – “GO FOR ORBIT” Documentary – https://www.youtube.com/watch?v=wlWF7kJokvk
Editorial
If you found this valuable, the I’d appreciate a ‘like’ or perhaps a donation at https://ko-fi.com/ciaops. This helps me know that people enjoy what I have created and provides resources to allow me to create more content. If you have any feedback or suggestions around this, I’m all ears. You can also find me via email director@ciaops.com and on X (Twitter) at https://www.twitter.com/directorcia.
If you want to be part of a dedicated Microsoft Cloud community with information and interactions daily, then consider becoming a CIAOPS Patron – www.ciaopspatron.com.
Watch out for the next CIA Brief next week
