OpenAI is highlighting 12 briefs arguing that training AI on copyrighted works can support research, accessibility, and new creative tools without handing users the original works as substitutes.

Source

Cybersecurity professionals are calling the rapid emergence of AI-detected vulnerabilities “Black October,” with some flaws appearing faster than organizations can patch them. Meanwhile, AI coding tools are helping attackers target banks and other institutions, raising fears that a major incident could arrive before the industry coordinates stronger safeguards.

Source

OpenAI, Google, Microsoft, and Anthropic will support a shared robots.txt signal that lets publishers state whether their content may be used to train AI models. The proposed policy token aims to replace separate opt-outs for each company’s crawler, but support will roll out gradually.

Source

A letter about Cardiff’s proposed data center has sharpened a question for its opponents: how much of their online activity depends on the facilities they oppose? It also points to a practical local concern—the site serves as a park-and-ride for the city’s main hospital.

Source

Blackstone expects Anthropic to receive a speculative-grade credit rating after its IPO, reversing its earlier forecast that the AI company would reach investment grade before 2027. The new forecast places Anthropic in the double-B range, the highest tier of junk-rated debt. A lower rating could make it harder and more expensive for the company to borrow as it funds its expanding computing capacity.

Source

Microsoft Execution Containers (MXC) sandbox AI agents, limiting which files and network resources they can access; process-container support spans Windows 11, macOS, and Linux. Microsoft is also expanding Project Perception with agent-specific identities, task-scoped permissions, and controls that can tighten access or stop an agent as its risk changes. The goal is to govern agents as distinct actors—not simply trust the user accounts they operate under.

Source

Microsoft Copilot brings AI assistance to Microsoft 365 apps, and its October update previews a cloud-based Autopilot agent that could keep work moving with less manual direction. The rollout also adds reusable skills for repeat tasks and new ways to organize projects and collaborate. Availability is staggered, with some features arriving in November or later.

Source

Microsoft says organizations should treat attackers probing OneDrive URLs for usernames as a low-risk threat, not a sign that accounts have been breached. The probes can reveal user principal names (UPNs), which attackers might later use for phishing, credential stuffing, or social engineering. Microsoft recommends strong identity protections to limit that downstream risk.

Source

Rufus, a Windows utility for creating bootable USB drives from operating-system images, has released beta version 4.16 with a workaround for an AMI UEFI NTFS driver bug. It also improves USB speed reporting, adding recognition of USB 3.2 (20 Gbps), USB4 (40 Gbps), and USB4 Version 2.0 (80 Gbps). The NTFS flaw can disrupt firmware checks of installation media on affected PCs.

Source

Only 13% of organizations believe they have adequate AI-agent governance, according to Gartner data cited by IBM. Microsoft’s guidance gives identity teams a practical starting point: strengthen authentication, remove standing privileges, and apply consistent access policies so agent identities receive least-privilege protections.

Source

Microsoft reports two intrusions that turned Teams help-desk impersonation and user-approved Quick Assist sessions into attacks on Active Directory Certificate Services (AD CS), which issues certificates used for authentication. One attacker abused an ESC1 template and continued using a privileged account excluded from automated containment to steal data. In a separate ESC8 attack, isolating the affected workstation stopped further observed activity.

Source

The Azure Developer CLI (azd) helps developers deploy applications from their local environment to Azure. Its October updates add a preview mode that shows deployment targets before changes are made, plus support for image-based deployments to Azure Container Apps express environments.

Source

Microsoft Intune manages macOS devices, policies, and apps. Its macOS deployment guide lays out the setup path, from prerequisites and enrollment to security controls and remote actions. It also distinguishes enrollment options for personal, corporate, and shared Macs.

Source

SharePoint Online now ships a knowledge base feature in which Copilot and AI agents assemble and maintain cross-linked Markdown pages from what a team already produces. Work IQ's knowledge compilation engine generates those pages; Microsoft asserts that agents can reply up to four times faster while using 30% fewer tokens when they read the curated pages rather than the underlying documents. The SharePoint knowledge base is in the Frontier preview only for now, and most AI operations depend on usage-based billing backed by Copilot credits. This piece covers how the capability works, what you need to evaluate it, and which preview limits touch permissions and content fidelity.

Source

Alibaba Cloud has opened its first public cloud region in the Netherlands, bringing two data centers online in Amsterdam. The region adds local infrastructure for customers seeking lower latency, disaster-recovery options, and data processing and storage within the Netherlands or Europe. Alibaba says its global footprint now spans 109 availability zones across 32 regions.

Source

Microsoft’s October 28 preview update, KB5127936, fixes Remote Desktop failures that blocked non-admin users on hosts with Administrator Protection enabled. The issue followed earlier updates and affected connections with error 0xC000000D; Microsoft’s fix builds on the resolution covered in its Windows 11 Remote Desktop error update.

Source

Claude Code’s VS Code extension brings Anthropic’s coding assistant to compatible editors. A tampered version, 2.1.313, was available through Cursor’s extension marketplace for about 11 hours on October 29, 2026, and attempted to install persistent SSH access on affected computers. Cursor says the attacker’s server is now offline, but affected users still need to check and clean their systems.

Source

Microsoft says the October 28 KB5127936 preview update finally fixes Windows 11 Always On VPN failures, after it had to retract its claim that October’s update resolved the bug. Always On VPN creates secure connections between devices and corporate networks; September’s security updates broke some profiles using automatic protocol selection or fallback between IKEv2 and SSTP. Affected connections could hang at “Connecting” or report that the specified port was already in use.

Source

Chrome Canary on Windows 11 now lets users defer an update restart until the screen stays on and unused for five minutes. Testing showed Chrome can restart and install the update while the user is away; Google has not announced when the feature will reach stable Chrome.

Source