Netcrook
AI Security & Agentic Systems
GhostApproval Turns AI Coding Assistants Into a File-System Trap
A trust-boundary flaw in six code agents shows how a malicious repository can manipulate sandboxed tools, not by breaking the model, but by abusing path handling and approval design.
When a PNG Becomes a Secret Courier for AI Agents
A demonstrated image-based prompt injection shows how a harmless-looking file can steer an AI coding workflow toward sensitive repository data.
AI Inside the Enterprise: The Security Risk Hiding in the Way Organizations Change
When artificial intelligence enters business workflows, the biggest weakness is often not the model itself but the gap between deployment, governance, training, and security.
When AI Writes Faster, Security Teams Pay Later
Cheap monthly subscriptions can hide a larger bill: the time, labor, and review work needed to make AI-assisted code safe enough to trust.
When AI Stops Being a License and Starts Being a Meter
Consumption-based AI pricing is turning token usage, model choice, and agent behavior into a finance problem that engineering teams can no longer ignore.
The Real Insider Threat May Wear a Token, Not a Badge
AI agents can stitch together harmless-looking permissions into a harmful outcome unless authorization, delegation, and audit are enforced at the moment each action happens.
Italy’s AI Rush Has a Human Problem: Governance Is Lagging Behind the Tools
The real question for businesses is no longer whether to adopt AI, but whether they can run it with clear ownership, oversight, and control.
Citrix Pushes NetScaler Into the MCP Layer of AI Traffic Control
Citrix has added MCP Gateway capabilities to NetScaler, giving enterprises a way to route, monitor, and control agent communications through a single policy point.
Citrix Puts AI Agent Traffic Under the Same Gate as Enterprise Apps
NetScaler’s new MCP Gateway features show how traditional traffic control is being refitted for agentic AI, where tool calls now need the same policy discipline as ordinary application flows.
When AI Starts Closing the Books, the Real Risk Moves to the Controls Room
AI agents can speed up financial close work in SAP-heavy environments, but the security question is no longer whether they can reconcile data - it is whether they can do so without weakening approvals, audit evidence, or release discipline.
HalluSquatting Shows How AI Assistants Can Be Tricked Into Pulling the Wrong Code
Researchers demonstrated a naming attack against AI assistants that can move from hallucinated lookups to remote code execution and, in some cases, malware delivery.
Why the Next AI Battleground Is Not the Model, but the Operating System Around It
The real advantage in enterprise AI may come from redesigning how organizations sense, decide, act, and learn, not from stacking more copilots on top of broken workflows.
Hiring by Algorithm: When Speed Turns Into a Quiet Filter for Bias
AI can make recruitment faster, but once software starts ranking people for rejection, the real risk is not efficiency - it is hidden exclusion, weak transparency, and decisions that are hard to challenge.
The Voice Machine Built for Care Also Defines a New Trust Problem
Voice for Purpose shows how AI can restore speech after illness or surgery, while also reminding defenders that any convincing voice model carries identity and fraud risks.
Why a Local AI Stack Matters When the Stakes Are Industrial
A new on-premises AI platform aimed at critical infrastructure is less about flashy model demos and more about where data lives, who controls the updates, and how much trust operators can actually place in automation.
When AI Starts Acting, Governance Stops Being Optional
Agentic systems can plan, use tools, consult data, and trigger workflows - but the productivity upside only survives if organizations can control every delegated action.
OpenAI Pushes AI Deeper Into Security Work - and Deeper Into Risk
GPT-5.6 and ChatGPT Work are built to do more cyber-relevant tasks, but the security story now hinges on how tightly their permissions, monitoring, and refusal layers hold up under pressure.
Government Chatbots Need More Than Good Manners: The Security Test Behind GenAI in Public Services
Public-sector assistants built with GenAI are only trustworthy if they survive adversarial testing, readable metrics, and expert review before citizens ever rely on them.
Microsoft Turns AI Loose on Windows Bugs, But the Real Weapon Is the Workflow
A multi-model agentic scanning harness is designed to find Windows flaws earlier, yet the security gain depends on validation, triage, and the patch pipeline behind it.
Datadog’s AI Push Turns Incident Response Into a Governed Machine
At Dash 2026, Datadog put Bits AI and its AI governance stack at the center of a strategy built on faster triage, tighter model control, and less blind trust in agentic systems.