Safeguard
Topic

AI Security

In-depth guides and analysis on ai security from the Safeguard engineering team.

676 articles

AI Security

Open-Weight Model Sandboxing Patterns

Running an open-weight model inside an enterprise perimeter seems safer than calling a hosted API. It is, and it isn't. The sandboxing patterns that actually produce the safety properties.

Aug 28, 20256 min read
AI Security

GPT-5 Launch: Reading the System Card for Supply-Chain Risk

GPT-5 shipped August 13, 2025 under OpenAI's Preparedness Framework v2. Here's what the system card tells security teams about deployment risk.

Aug 22, 20256 min read
AI Security

Local LLM Deployment: Enterprise Risks

Running LLMs on local hardware eliminates some risks and introduces others. A clear-eyed look at the enterprise risk profile of on-premise and on-device model deployments.

Aug 12, 20257 min read
AI Security

The Prompt Injection Problem Hiding Inside Everyday Code ...

AI coding assistants read untrusted files as instructions, not data. Here's how prompt injection sneaks malicious code into your commits — and how to catch it before it ships.

Aug 8, 20257 min read
AI Security

Model Training Data and the Propagation of Insecure Codin...

LLM coding assistants inherit insecure patterns from their training data — from SQLi-prone snippets to hallucinated packages attackers exploit. Here's how the risk propagates.

Aug 7, 20258 min read
AI Security

Securing AI Agents: MCP Protocol Risks and Mitigations

The Model Context Protocol is transforming how AI agents interact with tools, but it introduces new attack surfaces. Here is what security teams need to understand.

Jul 22, 20256 min read
AI Security

Tool Poisoning Attacks: How Malicious Instructions Hide I...

AI agent tools can hide invisible instructions attackers use to steal data. Here's how tool poisoning attacks work and how Safeguard stops them.

Jul 22, 20256 min read
AI Security

Model Context Protocol Security 101: What Could Go Wrong ...

MCP lets AI models call tools automatically — and lets malicious servers hide instructions in plain sight. Here's how tool poisoning, rug pulls, and shadowing actually work.

Jul 22, 20257 min read
AI Security

Agent Skill Marketplaces as the Next Frontier for Supply ...

Agent skill marketplaces are repeating npm and PyPI's supply chain mistakes—except the malicious payload is often a sentence of instructions, not code. Here's what's already been exploited.

Jul 22, 20257 min read
AI Security

Prompt Injection vs Traditional Injection Attacks: A Tech...

SQL injection was solved by separating code from data. Prompt injection can't be, because in an LLM they share one channel. Here's the technical comparison, with real exploits and dates.

Jul 22, 20258 min read
AI Security

Why Autonomous Coding Agents Need Their Own Threat Model

Coding agents run with real credentials and no pause button. Here is the threat model that treats them as autonomous infrastructure, not junior developers.

Jul 21, 20257 min read
AI Security

The Jailbreaking Economy: How Model Vulnerabilities Get D...

Jailbreak prompts now trade like exploits: sold as $200/month "dark" chatbots, bountied by vendors for up to $15,000. Here's how that market actually works.

Jul 21, 20258 min read
AI Security (Page 53) — Supply Chain Security Blog | Safeguard