<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
<title>Agentic AI in Cybersecurity</title>
<link>https://agenticaiincybersecurity.com/</link>
<description>What AI agents do in security work, job by job: what they read, what they do without a person, where a person approves and what evidence they leave.</description>
<language>en-US</language>
<atom:link href="https://agenticaiincybersecurity.com/feed.xml" rel="self" type="application/rss+xml"/>
<lastBuildDate>Thu, 19 Nov 2026 12:00:00 GMT</lastBuildDate>
<item><title>What Is an Agent Harness? Eight Definitions, One CVE</title><link>https://agenticaiincybersecurity.com/agent-harness/</link><guid>https://agenticaiincybersecurity.com/agent-harness/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Eight published agent harness definitions, quoted and dated, and a worked case: CVE-2026-82329 triaged by one model in two harnesses, with two results.</description></item>
<item><title>Agent Harness Approvals: Six Mechanisms, One Traced Call</title><link>https://agenticaiincybersecurity.com/agent-harness/approvals-and-checkpoints/</link><guid>https://agenticaiincybersecurity.com/agent-harness/approvals-and-checkpoints/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>How six agent harnesses pause a tool call for approval (versions read 19 Nov 2026), the OpenAI shell defaults, and a CVE-2026-0322 change traced to its log.</description></item>
<item><title>Agent Harness Memory: Six Designs and a Checklist File</title><link>https://agenticaiincybersecurity.com/agent-harness/context-and-memory/</link><guid>https://agenticaiincybersecurity.com/agent-harness/context-and-memory/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>How six agent harnesses keep memory (versions read 19 Nov 2026), plus a CVE-2026-9198 checklist file with agent and human fields and a retention table.</description></item>
<item><title>Agent Eval Harness: A Worked Security Example</title><link>https://agenticaiincybersecurity.com/agent-harness/evaluation-harness/</link><guid>https://agenticaiincybersecurity.com/agent-harness/evaluation-harness/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Eval harness for a triage agent: five tasks from CVE-2026-82329, CVE-2026-41709 and CVE-2026-9198, with graders, three trials each, and pass@3 against pass^3.</description></item>
<item><title>How to Build an Agent Harness: Steps, Owners, Records</title><link>https://agenticaiincybersecurity.com/agent-harness/how-to-build-an-agent-harness/</link><guid>https://agenticaiincybersecurity.com/agent-harness/how-to-build-an-agent-harness/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Eight steps to build a security agent harness, each with an owner, the record it leaves and how it fails, worked on a triage harness for CVE-2026-82329.</description></item>
<item><title>How to Test an Agent Harness: A Worked Test Run</title><link>https://agenticaiincybersecurity.com/agent-harness/how-to-test-an-agent-harness/</link><guid>https://agenticaiincybersecurity.com/agent-harness/how-to-test-an-agent-harness/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Eight harness tests tied to documented gaps in Claude Agent SDK 0.2.185, OpenAI Agents SDK 0.25.0 and Deep Agents 0.7.40, run on an Artifactory response agent.</description></item>
<item><title>Agent Harness Logging: A Filled Evidence Record</title><link>https://agenticaiincybersecurity.com/agent-harness/logging-and-evidence/</link><guid>https://agenticaiincybersecurity.com/agent-harness/logging-and-evidence/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>A filled evidence record for one agent action on Artifactory CVE-2026-82329, with the logging each SDK gives by default and why the log must live off the host.</description></item>
<item><title>One Agent or Many: Orchestration for Security Agents</title><link>https://agenticaiincybersecurity.com/agent-harness/orchestration-and-subagents/</link><guid>https://agenticaiincybersecurity.com/agent-harness/orchestration-and-subagents/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>When a security agent needs subagents: Slack's Director, Critic and four experts, a worked token bill at Anthropic's 4x and 15x figures, and SDK limits.</description></item>
<item><title>Tool Permissions for a Security Agent Harness</title><link>https://agenticaiincybersecurity.com/agent-harness/tools-and-permissions/</link><guid>https://agenticaiincybersecurity.com/agent-harness/tools-and-permissions/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>How six agent harnesses decide whether a tool call runs, five documented gaps to test, and a filled register for four security tools. Docs read 19 Nov 2026.</description></item>
<item><title>BountyBench Results: Detect, Exploit and Patch by Agent</title><link>https://agenticaiincybersecurity.com/benchmarks/bountybench/</link><guid>https://agenticaiincybersecurity.com/benchmarks/bountybench/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>BountyBench results for 10 agents on 120 tasks from 40 bug bounties, with model versions, attempts and bounty dollars, and how its invariant check works.</description></item>
<item><title>CTIBench Results: Threat Intel Tasks for LLMs</title><link>https://agenticaiincybersecurity.com/benchmarks/ctibench/</link><guid>https://agenticaiincybersecurity.com/benchmarks/ctibench/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>CTIBench (arXiv 2406.07599): 2,500 questions and four CTI tasks, with results for five models, sample sizes worked into intervals, and how teams use it.</description></item>
<item><title>CVE-Bench Results: Agents Exploiting 40 Web CVEs</title><link>https://agenticaiincybersecurity.com/benchmarks/cve-bench/</link><guid>https://agenticaiincybersecurity.com/benchmarks/cve-bench/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>CVE-Bench (arXiv 2503.17332): 40 critical web CVEs, eight graded outcomes, published success rates, token costs and failure modes for three 2025 agents.</description></item>
<item><title>CyberGym Results: Reproducing 1,507 Real Bugs</title><link>https://agenticaiincybersecurity.com/benchmarks/cybergym/</link><guid>https://agenticaiincybersecurity.com/benchmarks/cybergym/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>CyberGym (arXiv 2506.02548): agents reproduce 1,507 OSS-Fuzz bugs from a description. Four levels, the paper's baselines and the project table to Oct 2026.</description></item>
<item><title>SWE-bench Results: Dated Percentages and What They Mean</title><link>https://agenticaiincybersecurity.com/benchmarks/swe-bench/</link><guid>https://agenticaiincybersecurity.com/benchmarks/swe-bench/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>SWE-bench results from swebench.com on 19 Nov 2026: 13 models on Verified and Multilingual, Full history since 2023, and the Multimodal 480 vs 517 count.</description></item>
<item><title>Agent Harness vs Agent Framework: Which Layer to Pick</title><link>https://agenticaiincybersecurity.com/guides/agent-harness-vs-agent-framework/</link><guid>https://agenticaiincybersecurity.com/guides/agent-harness-vs-agent-framework/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Runtime, framework or harness for a security agent? One job, the Cisco ISE patch check, worked through four routes, with what the team must still configure.</description></item>
<item><title>How Long to Keep AI Agent Run Logs and Which Fields</title><link>https://agenticaiincybersecurity.com/guides/agent-run-logs-what-to-keep/</link><guid>https://agenticaiincybersecurity.com/guides/agent-run-logs-what-to-keep/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>EU AI Act Articles 19 and 26 set six months for high-risk systems from 2 Dec 2027. A keep, store or drop table per agent log field, plus a dated change record.</description></item>
<item><title>Agentic AI vs Generative AI in Security Work Compared</title><link>https://agenticaiincybersecurity.com/guides/agentic-ai-vs-generative-ai-in-security/</link><guid>https://agenticaiincybersecurity.com/guides/agentic-ai-vs-generative-ai-in-security/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>The same Atlassian advisory, CVE-2026-21589, handled as a one-shot summary and as an agent with tools: what changes in output, error cost and testing.</description></item>
<item><title>Agentic AI vs SOAR Playbooks: When a Playbook Is Enough</title><link>https://agenticaiincybersecurity.com/guides/agentic-ai-vs-soar-playbooks/</link><guid>https://agenticaiincybersecurity.com/guides/agentic-ai-vs-soar-playbooks/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>A SOAR playbook fixes its branches in advance; an agent picks its next step. On CVE-2026-0322: 20 fixed builds, one step for an agent, four for a playbook.</description></item>
<item><title>AI Agent Evals: Grading Runs and Setting a Release Gate</title><link>https://agenticaiincybersecurity.com/guides/ai-agent-evals/</link><guid>https://agenticaiincybersecurity.com/guides/ai-agent-evals/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Grade an agent eval with pass@1, pass@3 and pass^3 on 40 findings and 120 trials, then compare the result with a gate written before the run (19 Nov 2026).</description></item>
<item><title>AI Agent Guardrails: A Graded Tool List With Limits</title><link>https://agenticaiincybersecurity.com/guides/ai-agent-guardrails/</link><guid>https://agenticaiincybersecurity.com/guides/ai-agent-guardrails/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Build guardrails for a security agent as a graded tool list: nine tools for CVE-2026-111497 rated low, medium or high, each with access, undo and approval rule.</description></item>
<item><title>AI Agent Autonomy Levels: Rating Eight Security Actions</title><link>https://agenticaiincybersecurity.com/guides/autonomy-levels/</link><guid>https://agenticaiincybersecurity.com/guides/autonomy-levels/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Set autonomy per action, not per product: eight security actions rated on the NCSC's five dimensions, mapped to the five levels of Feng, McDonald and Zhang.</description></item>
<item><title>Build vs Buy AI Agents for Security: What Decides It</title><link>https://agenticaiincybersecurity.com/guides/build-vs-buy-ai-agents/</link><guid>https://agenticaiincybersecurity.com/guides/build-vs-buy-ai-agents/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Decide build or buy for a security agent on four tests before cost: a product for the job, an approval gate you set, exportable logs, a swappable model.</description></item>
<item><title>Choosing a Model for a Security Agent: Limits and Dates</title><link>https://agenticaiincybersecurity.com/guides/choosing-a-model-for-a-security-agent/</link><guid>https://agenticaiincybersecurity.com/guides/choosing-a-model-for-a-security-agent/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Context limits, retirement dates, retention terms and a worked 120-trial comparison of two models, read from Anthropic, Google and OpenAI pages on 19 Nov 2026.</description></item>
<item><title>Cost of Running Security Agents: Worked Per-Task Figures</title><link>https://agenticaiincybersecurity.com/guides/cost-of-running-security-agents/</link><guid>https://agenticaiincybersecurity.com/guides/cost-of-running-security-agents/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Token cost of two security tasks at twelve published price points, then caching, batch, retries and review time worked through. Prices read 19 Nov 2026.</description></item>
<item><title>Evaluate a Security Agent on Your Own Findings</title><link>https://agenticaiincybersecurity.com/guides/evaluate-a-security-agent-on-your-findings/</link><guid>https://agenticaiincybersecurity.com/guides/evaluate-a-security-agent-on-your-findings/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>A worked trial of a dependency-fix agent on 40 findings: four checks in a clean copy, a 52 to 80% interval on 27 passes, thresholds written first, the record.</description></item>
<item><title>How to Use Agentic AI in Cybersecurity: Six Steps</title><link>https://agenticaiincybersecurity.com/guides/how-to-use-agentic-ai-in-cybersecurity/</link><guid>https://agenticaiincybersecurity.com/guides/how-to-use-agentic-ai-in-cybersecurity/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Six steps to put an agent on one security job, with the owner, the record and the usual failure for each, NCSC risk levels and a dated rollout worked through.</description></item>
<item><title>Designing the Approval Step for a Security Agent</title><link>https://agenticaiincybersecurity.com/guides/human-approval-design/</link><guid>https://agenticaiincybersecurity.com/guides/human-approval-design/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>What an approver needs to judge an agent's change: a filled approval record for CVE-2026-75604, how many accepted items to sample, and expiry rules.</description></item>
<item><title>Human in the Loop vs Human on the Loop in Security</title><link>https://agenticaiincybersecurity.com/guides/human-in-the-loop-vs-human-on-the-loop/</link><guid>https://agenticaiincybersecurity.com/guides/human-in-the-loop-vs-human-on-the-loop/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>In the loop means a person approves before the action; on the loop means a person supervises after. Three actions from CVE-2026-123158 placed in each mode.</description></item>
<item><title>Measuring a Security Agent After It Ships</title><link>https://agenticaiincybersecurity.com/guides/measuring-an-agent-after-it-ships/</link><guid>https://agenticaiincybersecurity.com/guides/measuring-an-agent-after-it-ships/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Six production metrics for a security agent with formulas, thresholds, the action each triggers, a worked four-week review and Rival's block-rate series.</description></item>
<item><title>When a Security Task Needs More Than One Agent</title><link>https://agenticaiincybersecurity.com/guides/multi-agent-systems-in-security-work/</link><guid>https://agenticaiincybersecurity.com/guides/multi-agent-systems-in-security-work/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Split a security agent only for tool overload or privilege. OpenAI's 15 and 10 tool counts, LangChain's call and token counts, and a worked split of 14 tools.</description></item>
<item><title>What a Security Agent Should Never Do Alone</title><link>https://agenticaiincybersecurity.com/guides/what-an-agent-should-never-do-alone/</link><guid>https://agenticaiincybersecurity.com/guides/what-an-agent-should-never-do-alone/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Seven actions a security agent must not take without a named person, graded on the NCSC 0 to 4 scale with Red Hat's CVE-2026-46300 fixes and a real failure.</description></item>
<item><title>What Is Agentic AI in Cybersecurity? A Working Definition</title><link>https://agenticaiincybersecurity.com/guides/what-is-agentic-ai-in-cybersecurity/</link><guid>https://agenticaiincybersecurity.com/guides/what-is-agentic-ai-in-cybersecurity/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>Agentic AI in cybersecurity defined from Anthropic, OWASP and LangChain, then run on CVE-2026-46300 as a workflow and as an agent, with tests for a false label.</description></item>
<item><title>Writing a Task Spec for a Security Agent: Template</title><link>https://agenticaiincybersecurity.com/guides/writing-a-task-spec-for-a-security-agent/</link><guid>https://agenticaiincybersecurity.com/guides/writing-a-task-spec-for-a-security-agent/</guid><pubDate>Thu, 19 Nov 2026 12:00:00 GMT</pubDate><description>A complete eight-field task spec for a security agent, filled in for CVE-2026-46300 with real kernel versions, a 31 Oct 2026 due date and sourced stop rules.</description></item>
<item><title>Aikido reports its Deep Review flagged Java4Shell in the pull request that introduced it</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-18</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-18</guid><pubDate>Wed, 18 Nov 2026 12:00:00 GMT</pubDate><description>Aikido says its AI pull request review would have caught the OpenJDK Base64 decoder flaw (CVE-2026-114938) when the code change was first proposed, years before disclosure. The post also gives affected configurations and fixed Oracle releases.</description></item>
<item><title>Tenable adds a Tenable One import into OpenAI's Codex Security</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-18</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-18</guid><pubDate>Wed, 18 Nov 2026 12:00:00 GMT</pubDate><description>Tenable says selected Tenable One findings can be imported into the Codex Security workbench through a local CLI import, with Tenable's description and solution attached. The aim is to stop developers from re-investigating findings that lose context in ticket handoffs.</description></item>
<item><title>GitHub adds adaptive effort and secret detection to Copilot code review</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-17</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-17</guid><pubDate>Tue, 17 Nov 2026 12:00:00 GMT</pubDate><description>GitHub's changelog says Copilot code review can now choose a review depth per pull request based on complexity and risk, and flag exposed secrets in review comments. Both features are in public preview.</description></item>
<item><title>Endor Labs connects findings to OpenAI Codex for fix and verification</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-12</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-12</guid><pubDate>Thu, 12 Nov 2026 12:00:00 GMT</pubDate><description>Endor Labs says its findings now flow into Codex Security through OpenAI's Daybreak Defense Factory, carrying reachability and remediation guidance. In the described loop, Codex implements and tests a fix and Endor Labs checks the result against policy before human review.</description></item>
<item><title>Rival describes what happens when a model's cyber classifier flags security agents</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-12</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-12</guid><pubDate>Thu, 12 Nov 2026 12:00:00 GMT</pubDate><description>Rival reports that its package-analysis agents, which read vulnerable open-source code in a sandbox and judge exploitability, normally see about one thread in twenty blocked by a model's cyber safeguard. It measured the rate rising to about one in five on November 5.</description></item>
<item><title>Semgrep releases an MCP server so coding agents can triage and fix findings</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-12</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-12</guid><pubDate>Thu, 12 Nov 2026 12:00:00 GMT</pubDate><description>Semgrep released a beta MCP server that lets an agent search findings, triage in bulk, generate fixes and file tickets using the user's existing permissions. Semgrep states that the developer reviews each fix before it merges.</description></item>
<item><title>Endor Labs explains its detection, triage and remediation agents</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-11</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-11</guid><pubDate>Wed, 11 Nov 2026 12:00:00 GMT</pubDate><description>Endor Labs lays out a pipeline of three agents on top of deterministic program analysis. One detects candidates, one labels each finding true positive, false positive or unknown with written reasoning, and one proposes fixes as pull request comments for a developer to review.</description></item>
<item><title>Wiz describes how its Green Agent handles remediation bottlenecks</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-03</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-11-03</guid><pubDate>Tue, 03 Nov 2026 12:00:00 GMT</pubDate><description>Wiz describes three bottlenecks that slow cloud fixes: finding the owner, understanding the root cause, and getting the fix deployed. Its Green Agent suggests an assignee by analyzing commit history, service catalogs, CMDB data and cloud logs, so a security team can route an issue with one click.</description></item>
<item><title>Project Zero compares an LLM bug finder with a fuzzer on V8</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-27</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-27</guid><pubDate>Tue, 27 Oct 2026 12:00:00 GMT</pubDate><description>Project Zero compared Google's BigSleep, an agentic bug-finding system, with the Fuzzilli fuzzer on Chrome's V8 engine. Over six months BigSleep reported more bugs (45 versus 37), the overlap was partial, and the authors conclude the approaches complement each other.</description></item>
<item><title>Snyk opens its Evo Remediation Agent campaigns for preview</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-22</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-22</guid><pubDate>Thu, 22 Oct 2026 12:00:00 GMT</pubDate><description>Snyk released campaigns in open preview, where an AppSec team picks repositories and vulnerabilities and the agent opens pull requests with dependency upgrades. The agent works in a hosted sandbox, checks its work with builds, tests and rescans, and never merges; developers review and merge.</description></item>
<item><title>Snyk publishes VulnBump, a benchmark for AI agents fixing vulnerable dependencies</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-15</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-15</guid><pubDate>Thu, 15 Oct 2026 12:00:00 GMT</pubDate><description>Snyk tested agents on 307 real dependency upgrades, counting a pass only if the vulnerability was gone and the project still built and passed tests. It reports that the same model scored differently depending on whether it had scan data, and that an unaided agent often claimed fixes a rescan did not confirm. These are Snyk's own results.</description></item>
<item><title>Anthropic launches a Cyber Mission with open-source scanning</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-08</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-08</guid><pubDate>Thu, 08 Oct 2026 12:00:00 GMT</pubDate><description>Anthropic announced a program that includes OSS Scanner, which gives open-source projects regular free security scans by its models to find vulnerabilities and propose patches. It notes that verifying, prioritizing and fixing findings is still the hard part after Project Glasswing.</description></item>
<item><title>Anthropic expands its Cyber Verification Program into access tiers</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-06</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-10-06</guid><pubDate>Tue, 06 Oct 2026 12:00:00 GMT</pubDate><description>Anthropic merged Project Glasswing into an expanded Cyber Verification Program with three access tiers for vetted security teams. The Defense Access tier covers tasks like incident response, malware reverse engineering and validating vulnerabilities with reduced cyber blocking.</description></item>
<item><title>NCSC blog on why defenders struggle to use agents the way attackers do</title><link>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-09-21</link><guid>https://agenticaiincybersecurity.com/dates-to-know/#d-2026-09-21</guid><pubDate>Mon, 21 Sep 2026 12:00:00 GMT</pubDate><description>An NCSC blog by Dave Chismon argues that defense is limited mostly by organizational constraints such as patch windows and change approval, not by technical ones. It sets out what defenders can do to get value from agentic cyber defense.</description></item>
</channel>
</rss>
