An agent’s identity matters as much as its intelligence.
Who does it represent? Which tools can it call? What can it change? Who approves a high-impact action?
These answers define the trust boundary for enterprise AI.
The Agentic Endpoint introduces a new question for IT: What exactly is running on the device?
An AI agent has a model, instructions, tools, data access, and authority. Managing only its app package misses most of the story.
These are now enterprise management surfaces.
The Agentic Endpoint introduces a new question for IT: What exactly is running on the device?
An AI agent has a model, instructions, tools, data access, and authority. Managing only its app package misses most of the story.
These are now enterprise management surfaces.
For decades, we deployed apps to endpoints. Now, we’re deploying AI agents to the Agentic Endpoint. Agents that understand context, act with permission, and work alongside people.
For decades, we deployed apps to endpoints. Now, we’re deploying AI agents to the Agentic Endpoint. Agents that understand context, act with permission, and work alongside people.
The official MCP repos published 31 security advisories in 28 days this October. Almost none of them were "AI bugs."
A catalog of 44 MCP advisories found that 18 of them (41%) follow one pattern: untrusted text becomes a command. That text can be a URL, a config field, or a tool argument. Only two started with prompt injection. The rest were classic web flaws: missing auth, missing ownership checks, path traversal, and command injection.
The nastiest is an OAuth bug that appeared in the Python SDK, the TypeScript SDK, and FastMCP. Clients didn't bind credentials to a verified issuer, so a malicious server could steer refresh tokens to an attacker's authorization server. Upgrading alone doesn't fix it. You also have to set the issuer explicitly and re-bind any stored credentials. Separately, an MCP Inspector bug (CVSS 9.6) let a remote connection turn into a local process spawn.
As a former VC at Menlo Ventures, I'd treat this as diligence. Your agents' tool connections are a software supply chain. Boards should ask management for three things: an inventory of every MCP server the company's agents connect to, pinned minimum-safe versions, and a named approver for new servers.
One limitation: behavioral trust scoring and red-teaming (what TrustModel does) don't replace patching. You need AppSec for the plumbing and evals for the agent.
Who in your org owns the MCP server inventory?
Follow me here and on X (@karlmehta). #AISecurity #AIGovernance #MCP
@trustmodelai
‼️ Over 60,000 Attacks Target WordPress RCE Vulnerability Within 24 Hours
Details: https://t.co/30jqffFNSq
Hackers are targeting a WordPress Core remote code execution vulnerability that can expose websites during software updates. The supplied Wordfence threat intelligence snapshot lists 60,471 blocked attempts over 24 hours.
WordPress released version 7.1.4 on October 28, 2026, addressing 22 security vulnerabilities. The update closes the code execution weakness alongside stored cross-site scripting, information disclosure, and permission issues.
These are attack attempts, not confirmed website compromises, and the figure does not establish how many individual sites were targeted.
#CybersecurityNews
Faster patching is going to be one of biggest areas for growth in the agentic area. Companies are doing a reasonable job securing their networks and data centers but not as much focus has gone into securing the endpoint used by employees. Those endpoints will be more valuable attack vector for hackers as agentic use rises. @a16z had a pretty cool report showing speed of attack rising while the speed of patching remained slow (attaching it here https://t.co/8bMUPIo0ab). As an industry we need to much faster cycle time to remediate while giving the assurance that it won't break things. More on this soon :-)
For decades, endpoints ran software. Now they will run intelligence.
The next computing revolution won’t just happen in the cloud. It will happen at every endpoint.
Welcome to the era of Agentic Endpoints!
Autonomy starts with the admin. 👀
As AI agents take on more work, autonomous endpoint management (AEM) helps IT move faster without giving up precision, governance, or trust.
See how Omnissa is building the guardrails that make AEM possible. https://t.co/iJbdflLlR3
Fresh off the ONE Stage. 🎤🔥
We caught up with Hemant Sahani shortly after the #OmnissaONE keynote to get a quick recap of what he announced on the big stage today!
Miss the livestream? Watch the keynote on demand. https://t.co/1WyHTR5Tgs
“Now the announcement I am most excited about: Horizon Delegate. How many of you have seen someone carrying a laptop around the office, lid open, waiting for an AI tool to finish?” - Kristina De Nike, Director of Product Management, Omnissa
Garry Tan says we'll see a wave of exceptional older AI-native founders:
"There's going to be as many Patrick Collisons as ever, but one mega trend that we're seeing is the 35, 40, 45-year-old founder who's been around the block, built a lot of engineering."
"Peter Steinberger is a perfect example of that. He's been around the block. He knows what to build."
"If you take that person, suddenly there's 400 of those people. You can outperform an entire department of any Mag 7."
@garrytan@illscience
A 1,000% increase in workplace AI app usage tells us one thing.
Employees aren’t waiting for AI strategies. They’re already using AI.
The conversation now is how to enable it securely.
Great discussion with @NeilCHughes.
🎙️ https://t.co/wNEcFJhBO3
Heading to #NRF2025? We are too! 🤩
Stop by booth #545 for live demos of The Omnissa Platform, including @WorkspaceONE and @HorizonVDI. Let’s shape the future of #retail together! https://t.co/TesU3j9Sfz
Big news! Omnissa is partnering with @CrowdStrike to connect the digital workspace with endpoint protection and extended detection and response solutions. Together, we bring AI-native cybersecurity and optimized productivity. 🤝 Learn more. https://t.co/hyGnrudFc7