A hallucination is a generated statement that appears plausible but is unsupported or incorrect. No single prompt eliminates the problem. Reliable applications reduce hallucination risk through architecture, evidence, validation, and user-experience decisions. Start by limiting the task Open-ended requests encourage the model to draw broadly on learned patterns. Narrow tasks with explicit scope and known information boundaries are easier to verify. Generative models are designed to produce plausible sequences, not guaranteed facts; AI-900 concepts overview provides the workload foundation for understanding why fluency and factual certainty are different properties….

  Generative and agentic applications introduce security risks that appear when natural-language inputs influence models, retrieval, tools, and actions. The main question is not whether a model can be perfectly trusted; it is whether the surrounding system can constrain what untrusted input is allowed to cause. Treat prompts and retrieved text as untrusted input User messages, uploaded documents, web content, and retrieved knowledge may contain instructions that conflict with application policy. Prompt injection occurs when such content changes model behavior in an unintended way. AI security has to cover the…

  AI governance defines how an organization decides which AI uses are acceptable, who is accountable, how risks are assessed, and what evidence is required before and after deployment. It is a management system around AI, not a substitute for technical security controls. Begin with an AI inventory Organizations cannot govern systems they do not know exist. Maintain an inventory of AI applications, owners, purpose, models or providers, data categories, users, and important dependencies. Governance discussions become more precise when teams share the workload vocabulary in AI-900 concepts foundation before…

Choosing an AI model in 2026 is no longer a contest to find the model with the highest benchmark score. Production teams are balancing quality, latency, throughput, context size, tool use, privacy, deployment control, regional availability, safety, observability, and cost. A model that looks strongest in a public leaderboard can be the wrong choice for a customer-support workflow, an offline industrial application, a regulated knowledge assistant, or a high-volume classification pipeline. The most useful starting point is to separate the model decision from the application decision. The application includes retrieval,…

  Secrets management protects the credentials that applications, automation, administrators, and services use to authenticate or establish trust. API keys, passwords, tokens, certificates, and private keys become dangerous when they are copied into source code, shared through chat, stored in plaintext configuration, or left active long after the original owner or workload has changed. A strong secrets program reduces how many secrets exist, where they can be stored, who can retrieve them, and how long they remain valid. Start by identifying what is actually a secret Secrets include database passwords,…

Data loss prevention is often reduced to a blocking rule—stop users from emailing a sensitive file. Real DLP programs are broader. They discover sensitive information, classify it, understand where it moves, define policy by context, monitor behavior, prevent or warn when appropriate, investigate incidents, and feed lessons back into data handling and access design. The difficulty is balancing protection with legitimate work. A rule that blocks every transfer containing a credit-card pattern may produce so many false positives that users route around it. A rule that only logs activity may…

  Security metrics are useful when they help leaders and practitioners decide whether risk is decreasing, controls are working, and important work is being completed. They become harmful when teams optimize for easy numbers such as alert volume, training completion, or vulnerability count without explaining what those numbers mean. A strong measurement program connects metrics to decisions, ownership, and evidence. Begin with the decision the metric supports Before creating a dashboard, ask who will use the measure and what decision it should influence. An SOC manager may need queue age…

  Security architecture patterns are reusable ways of organizing controls around common risks. They are not vendor products and they are not formulas that guarantee security. Their value comes from making trust assumptions explicit, limiting blast radius, separating responsibilities, and ensuring that failure of one control does not automatically expose the entire system. Defense in depth assumes controls can fail A layered design uses multiple controls that protect different points in the attack path. Authentication, authorization, segmentation, endpoint protection, encryption, logging, and recovery each address different failure modes. Security architecture…

  Security logging is the deliberate collection of evidence that can explain important actions across identities, endpoints, networks, cloud platforms, applications, and data systems. Collecting everything forever is rarely practical. The useful goal is to preserve the events needed to detect risky behavior, investigate incidents, validate controls, and satisfy audit requirements without drowning analysts in redundant noise. Begin with security questions Ask what the organization may need to prove later. Who authenticated? Which administrator changed policy? What process executed? Which system contacted an external destination? Who accessed sensitive data? The…

  Networking connects users, devices, applications, cloud services, and data. Network security controls who can communicate, which paths are allowed, and what evidence is available when something goes wrong. This hub organizes the core ideas learners need across protocol fundamentals, switching, routing, wireless, WAN, cloud networking, security, troubleshooting, and certification study. Start with network fundamentals Before advanced design, learners need a dependable model of addressing, subnetting, gateways, ports, protocols, frames, packets, and routing decisions. Network+ fundamentals provides that entry-level foundation for the deeper topics in this hub. Strong fundamentals make…

  The OSI and TCP/IP models are frameworks for understanding how network communication is divided into layers. They are most useful as troubleshooting and learning tools: they help separate physical connectivity, local delivery, routing, transport, and application behavior so that one symptom can be investigated methodically. Modern protocols do not always fit perfectly into one box, but the layered approach remains valuable. Why networking uses layers Layers let technologies evolve without every component needing to understand every detail below or above it. An application can send data without knowing how…

  IPv4 subnetting is easier when it is treated as a design and troubleshooting skill rather than a memorization exercise. A prefix defines which bits describe the network and which remain available for hosts. From that one decision come the address range, broadcast boundary, usable space, route summarization opportunities, and many of the security and operational boundaries administrators work with every day. Begin with the prefix, not the dotted mask CIDR notation such as /24 or /27 states how many bits belong to the network prefix. The remaining bits form…

  Routing is the process of choosing where packets should go when the destination is not on the local network. Routers build a forwarding view from connected networks, static configuration, and routes learned through protocols. The durable skill is understanding how a device chooses among candidates, what makes a route usable, and how the network reacts when topology changes. A route describes a destination and a next step A routing-table entry normally includes a destination prefix and information about how to reach it. That may be a directly connected interface,…

  Ethernet switching moves frames inside a local Layer 2 domain. Modern switches learn where MAC addresses are reachable, separate traffic into VLANs, carry multiple VLANs across trunks, and use loop-prevention mechanisms so redundant links do not create uncontrolled frame circulation. Good switching design balances simplicity, failure tolerance, broadcast scope, and the need to route between segments. Switches learn from source addresses When a frame arrives, a switch observes the source MAC address and associates it with the incoming interface in the relevant VLAN. For a known destination, the frame…

DNS, DHCP, and NAT solve different networking problems, but administrators often troubleshoot them together because a failure in one can look like a failure in another. DHCP gives a client usable IP configuration. DNS translates names into addresses and supports service discovery. NAT changes address information as traffic crosses a boundary. When users say “the network is down,” understanding those roles helps isolate the real layer of failure. These services are foundational because applications depend on them quietly. A laptop can have a working wireless link yet fail because it…