How these articles are researched and reviewed

Who writes them

I write and am accountable for every guide and insight on this site. Each one is checked and approved by me before it is published, and carries the date it was published and the date it was last reviewed.

I use AI tools to search the research literature and to help draft and edit. They do not decide what an article claims: I check every factual statement against the source it cites. The illustrations that accompany articles are generated with AI tools.

Where the evidence comes from

The articles draw on peer-reviewed research and preprints in AI security, which I maintain as a searchable library for my own research. Claims about attacks and defences link to their sources. Preprints are marked as such, because they have not yet been through peer review, and results are reported with the conditions under which they were measured.

Where the evidence is thin, contested or only from laboratory benchmarks, the article says so.

What the articles are, and are not

They explain how AI systems are attacked and how to design controls around them. They are general information, not an assessment of your systems and not legal advice. Security depends on your architecture, data and threat model.

No article is sponsored, and I am not paid by any vendor whose product is mentioned.

Corrections and updates

I review articles as the research moves, and at least every six months. Substantive changes are listed in an "Updates" note at the end of the article, with the date. If you find an error, please email contact@brentcoker.com.