AI Security
In-depth guides and analysis on ai security from the Safeguard engineering team.
676 articles
Griffin AI vs Mistral Large for Remediation
Mistral Large is a strong reasoning model, but remediation is more than generating a diff. We look at what Griffin AI adds for production fix workflows.
SSO & SCIM: Griffin AI vs Mythos
Enterprise identity is not a paywall. It is the substrate on which every other security control depends, and it is where Mythos-class vendors quietly fall behind.
The MCP Threat Model: What Actually Matters in 2026
Most MCP threat models confuse protocol risk with deployment risk. Here is what the real attack surface looks like after a year of production incidents.
SWE-Bench With Security Extensions: Field Review
SWE-bench became the default benchmark for measuring AI coding agents, but the security extensions that were bolted on afterwards deserve their own scrutiny. A field review of what they measure, where they break, and whether you should trust the numbers.
Griffin AI vs Claude Opus for Triage
Griffin uses Claude Opus as its deepest reasoning engine. Here's what triage looks like with Opus alone versus Opus running inside Griffin's eval harness.
Auto-Fix Compile Rates: Griffin AI vs Mythos
Griffin AI's auto-fixes compile clean 73 percent of the time and pass with minor edits 87 percent. Mythos-class pure-LLM patches rarely show those numbers for a reason.
Fine-Tuning Security LLMs vs Grounding: Which Wins
Fine-tuning teaches a model to be a security expert. Grounding lets a general model act like one by reading the right sources. The right answer is usually both, but the proportions matter.
EU CRA Readiness: Griffin AI vs Mythos
The EU Cyber Resilience Act wants mandatory vulnerability handling, SBOM delivery, and documented due diligence. Griffin AI produces those artifacts continuously. Mythos-class tools produce conversations about them.
Griffin AI vs OpenAI Codex for Security
Codex-style coding agents are powerful for writing features. Security remediation needs a different shape of system—one that grounds frontier reasoning in SBOM, policy, and reachability context.
Anthropic MCP Security Model: A Deep Dive
Anthropic's Model Context Protocol introduces a new trust boundary between agents and tools. Here is how the security model actually works in practice.
CycloneDX ML-BOM in 1.7: Implementation Guide
CycloneDX 1.7 was published in October 2025 and adopted by the General Assembly in December. We unpack what the ML-BOM capability means in practice for AI inventory.
Context Window As A Security Limit
The context window is usually marketed as a capability parameter. In a security setting, it behaves like a budget, a forgetting function, and an attack surface all at once.