Normal view

Received yesterday — 13 February 2026

Securing Agentic AI Connectivity

12 February 2026 at 17:50

 

Securing Agentic AI Connectivity

AI agents are no longer theoretical, they are here, powerful, and being connected to business systems in ways that introduce cybersecurity risks! They’re calling APIs, invoking MCPs, reasoning across systems, and acting autonomously in production environments, right now.

And here’s the problem nobody has solved: identity and access controls tell you WHO is acting, but not WHY.

An AI agent can be fully authenticated, fully authorized, and still be completely misaligned with the intent that justified its access. That’s not a failure of your tools. That’s a gap in the entire security model.

This is the problem ArmorIQ was built to solve.

ArmorIQ secures agentic AI at the intent layer, where it actually matters:

· Intent-Bound Execution: Every agent action must trace back to an explicit, bounded plan. If the reasoning drifts, trust is revoked in real time.

· Scoped Delegation Controls: When agents delegate to other agents or invoke tools via MCPs and APIs, authority is constrained and temporary. No inherited trust. No implicit permissions.

· Purpose-Aware Governance: Access isn’t just granted and forgotten. It expires when intent expires. Trust is situational, not permanent.

If you’re a CISO, security architect, or board leader navigating agentic AI risk — this is worth your attention.

See what ArmorIQ is building: https://armoriq.io

The post Securing Agentic AI Connectivity appeared first on Security Boulevard.

Received before yesterday

We let Chrome's Auto Browse agent surf the web for us—here's what happened

12 February 2026 at 07:00

We are now a few years into the AI revolution, and talk has shifted from who has the best chatbot to whose AI agent can do the most things on your behalf. Unfortunately, AI agents are still rough around the edges, so tasking them with anything important is not a great idea. OpenAI launched its Atlas agent late last year, which we found to be modestly useful, and now it's Google's turn.

Unlike the OpenAI agent, Google's new Auto Browse agent has extraordinary reach because it's part of Chrome, the world's most popular browser by a wide margin. Google began rolling out Auto Browse (in preview) earlier this month to AI Pro and AI Ultra subscribers, allowing them to send the agent across the web to complete tasks.

I've taken Chrome's agent for a spin to see whether you can trust it to handle tedious online work for you. For each test, I lay out the problem I need to solve, how I prompted the robot, and how well (or not) it handled the job.

Read full article

Comments

© Aurich Lawson

Xcode 26.3 adds support for Claude, Codex, and other agentic tools via MCP

3 February 2026 at 13:01

Apple has announced a new version of Xcode, the latest version of its integrated development environment (IDE) for building software for its own platforms, like the iPhone and Mac. The key feature of 26.3 is support for full-fledged agentic coding tools, like OpenAI's Codex or Claude Agent, with a side panel interface for assigning tasks to agents with prompts and tracking their progress and changes.

This is achieved via Model Context Protocol (MCP), an open protocol that lets AI agents work with external tools and structured resources. Xcode acts as an MCP endpoint that exposes a bunch of machine-invocable interfaces and gives AI tools like Codex or Claude Agent access to a wide range of IDE primitives like file graph, docs search, project settings, and so on. While AI chat and workflows were supported in Xcode before, this release gives them much deeper access to the features and capabilities of Xcode.

This approach is notable because it means that even though OpenAI and Anthropic's model integrations are privileged with a dedicated spot in Xcode's settings, it's possible to connect other tooling that supports MCP, which also allows doing some of this with models running locally.

Read full article

Comments

© Apple

Google begins rolling out Chrome's "Auto Browse" AI agent today

28 January 2026 at 13:00

Google began stuffing Gemini into its dominant Chrome browser several months ago, and today the AI is expanding its capabilities considerably. Google says the chatbot will be easier to access and connect to more Google services, but the biggest change is the addition of Google's autonomous browsing agent, which it has dubbed Auto Browse. Similar to tools like OpenAI Atlas, Auto Browse can handle tedious tasks in Chrome so you don't have to.

The newly unveiled Gemini features in Chrome are accessible from the omnipresent AI button that has been lurking at the top of the window for the last few months. Initially, that button only opened Gemini in a pop-up window, but Google now says it will default to a split-screen or "Sidepanel" view. Google confirmed the update began rolling out over the past week, so you may already have it.

You can still pop Gemini out into a floating window, but the split-view gives Gemini more room to breathe while manipulating a page with AI. This is also helpful when calling other apps in the Chrome implementation of Gemini. The chatbot can now access Gmail, Calendar, YouTube, Maps, Google Shopping, and Google Flights right from the Chrome window. Google technically added this feature around the middle of January, but it's only talking about it now.

Read full article

Comments

© Google

Microsoft Patch Tuesday, November 2025 Edition

16 November 2025 at 16:47

Microsoft this week pushed security updates to fix more than 60 vulnerabilities in its Windows operating systems and supported software, including at least one zero-day bug that is already being exploited. Microsoft also fixed a glitch that prevented some Windows 10 users from taking advantage of an extra year of security updates, which is nice because the zero-day flaw and other critical weaknesses affect all versions of Windows, including Windows 10.

Affected products this month include the Windows OS, Office, SharePoint, SQL Server, Visual Studio, GitHub Copilot, and Azure Monitor Agent. The zero-day threat concerns a memory corruption bug deep in the Windows innards called CVE-2025-62215. Despite the flaw’s zero-day status, Microsoft has assigned it an “important” rating rather than critical, because exploiting it requires an attacker to already have access to the target’s device.

“These types of vulnerabilities are often exploited as part of a more complex attack chain,” said Johannes Ullrich, dean of research for the SANS Technology Institute. “However, exploiting this specific vulnerability is likely to be relatively straightforward, given the existence of prior similar vulnerabilities.”

Ben McCarthy, lead cybersecurity engineer at Immersive, called attention to CVE-2025-60274, a critical weakness in a core Windows graphic component (GDI+) that is used by a massive number of applications, including Microsoft Office, web servers processing images, and countless third-party applications.

“The patch for this should be an organization’s highest priority,” McCarthy said. “While Microsoft assesses this as ‘Exploitation Less Likely,’ a 9.8-rated flaw in a ubiquitous library like GDI+ is a critical risk.”

Microsoft patched a critical bug in OfficeCVE-2025-62199 — that can lead to remote code execution on a Windows system. Alex Vovk, CEO and co-founder of Action1, said this Office flaw is a high priority because it is low complexity, needs no privileges, and can be exploited just by viewing a booby-trapped message in the Preview Pane.

Many of the more concerning bugs addressed by Microsoft this month affect Windows 10, an operating system that Microsoft officially ceased supporting with patches last month. As that deadline rolled around, however, Microsoft began offering Windows 10 users an extra year of free updates, so long as they register their PC to an active Microsoft account.

Judging from the comments on last month’s Patch Tuesday post, that registration worked for a lot of Windows 10 users, but some readers reported the option for an extra year of updates was never offered. Nick Carroll, cyber incident response manager at Nightwing, notes that Microsoft has recently released an out-of-band update to address issues when trying to enroll in the Windows 10 Consumer Extended Security Update program.

“If you plan to participate in the program, make sure you update and install KB5071959 to address the enrollment issues,” Carroll said. “After that is installed, users should be able to install other updates such as today’s KB5068781 which is the latest update to Windows 10.”

Chris Goettl at Ivanti notes that in addition to Microsoft updates today, third-party updates from Adobe and Mozilla have already been released. Also, an update for Google Chrome is expected soon, which means Edge will also be in need of its own update.

The SANS Internet Storm Center has a clickable breakdown of each individual fix from Microsoft, indexed by severity and CVSS score. Enterprise Windows admins involved in testing patches before rolling them out should keep an eye on askwoody.com, which often has the skinny on any updates gone awry.

As always, please don’t neglect to back up your data (if not your entire system) at regular intervals, and feel free to sound off in the comments if you experience problems installing any of these fixes.

[Author’s note: This post was intended to appear on the homepage on Tuesday, Nov. 11. I’m still not sure how it happened, but somehow this story failed to publish that day. My apologies for the oversight.]

❌