Blog
Insights from our experts on deploying AI Agents securely and effectively)
BSI AI Security Guidance for Firms
BSI AI security guidance is voluntary: C5, AIC4, and 2025 generative AI briefs for German enterprises, read alongside the binding EU AI Act.
)
)
AI Gateway PoC UK CISO Checklist
An AI Gateway PoC UK CISO should test enforcement, audit evidence, and data handling, not a slide demo. A 2026 checklist drawn from ICO and NCSC guidance.
)
UK Data Residency AI Gateway Guide
UK data residency AI gateway buyers should ask who can access prompts, logs, and embeddings, not only where servers sit. ICO and NCSC tests for 2026.
)
OpenCode vs Claude Code: 2026 Comparison
OpenCode vs Claude Code compared on benchmarks, pricing, permissions and security, including OpenCode's 2026 CVEs and Console breach.
)
AI Model Security: An Enterprise Guide
AI model security covers model files, training data, weights, and runtime. See the 2026 data on malicious models, which controls work, and where to start.
)
AI Agent Security: Distillation Attacks
AI agent security now covers distillation attacks that copy your model. See how they work, what the 2026 cases show, and the controls that cut extraction.
)
MCP Gateway Sidecar Pattern for Kubernetes
An MCP gateway sidecar puts auth, tool policy, and audit logs in every agent pod. See how the Kubernetes pattern works, its trade-offs, and when to use it.
)
GPT-6 Cyber: Access & Controls
GPT-6 Cyber governance for CISOs: Daybreak Red access, OpenAI safeguards vs your agent runtime controls for tools, network, credentials and logs.
)
Claude Code Plugins: Guide and Security
Claude Code plugin guide: what plugins contain, how to install them, official marketplace data, 2026 incidents and the controls to set first.
)
Claude Connectors: Setup and Security
Claude connectors explained: how to add directory and custom MCP connectors, the best ones to start with, admin controls and security risks.
)
Copilot CLI Prompt Injection Attack Chain
Copilot CLI prompt injection with encrypted payloads can leak .env files in autopilot. See the chain, two 2026 CVEs and the controls that still work.
)
Google ADK Security: A Production Guide
Google ADK security in production: the six 2026 CVEs to patch, callbacks, plugins, tool confirmation, MCP tool filters and A2A trust.
)
AgentCore Security: Gaps You Still Own
AgentCore security in 2026: what AWS secures, what you still own across Runtime, Gateway, Policy and Code Interpreter, and the CVEs to patch.
)
OpenClaw Skills: Setup, Picks and Risks
OpenClaw skills explained: the SKILL.md format, ClawHub install commands, a safe-install checklist and the 2026 malicious skill campaigns.
)
AGENTS.md: Guide, Risks and Security
AGENTS.md tells coding agents how to work in your repo. Learn to write one, why a planted file hijacked Codex in tests, and how to lock it down.
)
MCP Memory Server: Add Memory to Claude
Connect an MCP memory server to Claude Code with one command so your agent remembers across chats. Setup steps, file paths and security checks.
)
MCP Audit Logging: What CISOs Must Track
MCP audit logging for CISOs: what to log per tool call, how to feed your SIEM, EU AI Act, SOC 2 and ISO 42001 evidence, and shadow MCP detection.
)
Claude Code MCP and Cursor MCP Security
Claude Code MCP and Cursor MCP can run code with your credentials. Learn the config files, real CVEs and the admin controls that lock both down.
)
MCP Tool Poisoning and Rug Pull Attacks
MCP tool poisoning in 2026: how description injection, rug pulls and tool shadowing work, what new research shows, and how to stop them.
)
MCP Security Best Practices Checklist
MCP security best practices for 2026: a checklist for servers and clients on least privilege, tool pinning, sandboxing, OAuth and runtime enforcement.
AI Gateway PoC UK CISO Checklist
An AI Gateway PoC UK CISO should test enforcement, audit evidence, and data handling, not a slide demo. A 2026 checklist drawn from ICO and NCSC guidance.
UK Data Residency AI Gateway Guide
UK data residency AI gateway buyers should ask who can access prompts, logs, and embeddings, not only where servers sit. ICO and NCSC tests for 2026.
OpenCode vs Claude Code: 2026 Comparison
OpenCode vs Claude Code compared on benchmarks, pricing, permissions and security, including OpenCode's 2026 CVEs and Console breach.
AI Model Security: An Enterprise Guide
AI model security covers model files, training data, weights, and runtime. See the 2026 data on malicious models, which controls work, and where to start.
AI Agent Security: Distillation Attacks
AI agent security now covers distillation attacks that copy your model. See how they work, what the 2026 cases show, and the controls that cut extraction.
MCP Gateway Sidecar Pattern for Kubernetes
An MCP gateway sidecar puts auth, tool policy, and audit logs in every agent pod. See how the Kubernetes pattern works, its trade-offs, and when to use it.
GPT-6 Cyber: Access & Controls
GPT-6 Cyber governance for CISOs: Daybreak Red access, OpenAI safeguards vs your agent runtime controls for tools, network, credentials and logs.
Claude Code Plugins: Guide and Security
Claude Code plugin guide: what plugins contain, how to install them, official marketplace data, 2026 incidents and the controls to set first.
Claude Connectors: Setup and Security
Claude connectors explained: how to add directory and custom MCP connectors, the best ones to start with, admin controls and security risks.
Copilot CLI Prompt Injection Attack Chain
Copilot CLI prompt injection with encrypted payloads can leak .env files in autopilot. See the chain, two 2026 CVEs and the controls that still work.
Google ADK Security: A Production Guide
Google ADK security in production: the six 2026 CVEs to patch, callbacks, plugins, tool confirmation, MCP tool filters and A2A trust.
AgentCore Security: Gaps You Still Own
AgentCore security in 2026: what AWS secures, what you still own across Runtime, Gateway, Policy and Code Interpreter, and the CVEs to patch.
OpenClaw Skills: Setup, Picks and Risks
OpenClaw skills explained: the SKILL.md format, ClawHub install commands, a safe-install checklist and the 2026 malicious skill campaigns.
AGENTS.md: Guide, Risks and Security
AGENTS.md tells coding agents how to work in your repo. Learn to write one, why a planted file hijacked Codex in tests, and how to lock it down.
MCP Memory Server: Add Memory to Claude
Connect an MCP memory server to Claude Code with one command so your agent remembers across chats. Setup steps, file paths and security checks.
MCP Audit Logging: What CISOs Must Track
MCP audit logging for CISOs: what to log per tool call, how to feed your SIEM, EU AI Act, SOC 2 and ISO 42001 evidence, and shadow MCP detection.
Claude Code MCP and Cursor MCP Security
Claude Code MCP and Cursor MCP can run code with your credentials. Learn the config files, real CVEs and the admin controls that lock both down.
MCP Tool Poisoning and Rug Pull Attacks
MCP tool poisoning in 2026: how description injection, rug pulls and tool shadowing work, what new research shows, and how to stop them.
MCP Security Best Practices Checklist
MCP security best practices for 2026: a checklist for servers and clients on least privilege, tool pinning, sandboxing, OAuth and runtime enforcement.