AI Security
Griffin AI vs Fine-Tuned Open Weights for SecOps
Fine-tuning an open-weight model sounds like a shortcut to a custom SecOps copilot. In practice, it is one step of a much longer journey.
Feb 19, 20267 min read
Deep dives, practical guides, and incident analyses from engineers who build Safeguard. No fluff, no vendor FUD — just what you need to ship secure software.
Fine-tuning an open-weight model sounds like a shortcut to a custom SecOps copilot. In practice, it is one step of a much longer journey.
A benchmark that the model has seen in training is a benchmark of memorisation. Specific leakage-testing methods separate generalisation from recall.
Claude Desktop's MCP support makes it a capable security tool. Griffin AI builds on that foundation rather than competing with it.
An architectural comparison of Griffin AI's engine-grounded reasoning stack against the pure-LLM pattern that Mythos-class products rely on.
Multi-modal models bring image, audio, and video into the AI supply chain. Each modality introduces provenance and integrity challenges that text-only pipelines never had to face.
LLM-generated Dockerfiles repeat the same six or seven mistakes. Here is the pattern catalog and how to catch them before they ship.
Slopsquatting is the practice of registering package names that LLMs hallucinate, turning AI coding assistants into an accidental distribution channel.
Function calling gives models the ability to act. Acting safely on behalf of a specific user, in a specific context, within specific policy is a different problem.
LLM selection is ultimately a cost-quality optimisation under workflow constraints. The curve is not smooth, and the right point on it depends on where errors land in your pipeline.
Weekly insights on software supply chain security, delivered to your inbox.