How we use AI

Malwlab's reader-facing content is produced through an AI/LLM workflow. Drafting, research support, and editing all run through language models, and we mark that assistance plainly on every post where it is substantial.

The model does the text work. It drafts the first pass, pulls together source material, and helps with editing. It does not invent lab results, testing data, or personal observations. Where a claim comes from our own homelab, it works from our notes, not from imagination.

The human stays in the pipeline. Every AI-assisted post goes through human editorial review before it ships. We check technical claims against the source material, verify the numbers, and make the final editorial calls. If a claim does not trace back to a source or to our own work, it gets cut or corrected.

We hold three standards:

• Quality gates: a post does not publish until it passes editorial review. Failing checks route it back for a fix, and any change requires a fresh audit.
• Responsible source handling: external sources are cited with links where possible. Claims that are not our own firsthand work are attributed. Uncertainty and assumptions are stated explicitly, and when a claim draws on several sources, each one is named.
• No fabricated citations: we do not invent references, advisories, or benchmark results. A citation either points to something real and checkable, or it is not in the post.

The broader research approach, how we handle IOCs and technical detail, and how we track corrections live on our Methodology page. This page covers only the AI/LLM workflow.

Content with substantial AI assistance is tagged AI/LLM Generated.