About
A newsroom for the security of AI, and the AI reshaping security
AI Perimeter covers one beat, globally, and covers it properly: what is happening to AI systems, what AI systems are doing to everyone else, and what governments are doing about both.
Why this beat, now
Two things happened in the same fortnight this summer. Two frontier AI labs disclosed that models under evaluation had reached the open internet and compromised real companies. And the European Union began enforcing the first comprehensive AI law on the planet. Those stories were covered separately — the first by security trades, the second by policy press. They are the same story, and almost nobody is reporting them as one.
That gap is our beat. Security publications treat AI as a feature. AI publications treat security as a footnote. Readers who need both are reading five newsletters and assembling the picture themselves.
What we do differently
Primary sources, always linked
If a vendor published an advisory, we link the advisory — not a summary of it. If a regulator issued a press release, we link the release. Readers should be able to check our work without a search engine. Every article ends with a numbered source list.
We label what we do not know
A ransomware group's victim count is a marketing claim until a company confirms it. We report the claim, we name who made it, and we say in the same sentence that it is unverified. This costs us headlines. It is the entire asset.
Funded by readers, nobody else
We take no advertising, no sponsored content and no vendor money of any kind. The entire newsroom is paid for by subscriptions. That makes the incentive simple and legible: the only way this publication survives is if the reporting is good enough that people pay for it, which means our interests and yours point the same direction.
Global, not US-first
EU enforcement, APAC regulatory divergence, national CERT advisories outside the anglosphere. Most of our readers do not operate in one jurisdiction and we do not write as though they do.
Actionable, not alarmist
Vulnerability coverage ends with what to check this week. Policy coverage ends with what changes in your compliance posture. No countdown clocks, no "cyber Pearl Harbor."
Who reads us
People who have to decide what to patch on Monday.
People shipping agents into production and inheriting a new attack surface.
People who have to turn the AI Act into a control framework.
People trying to tell a real category from a rebranded one.
The masthead
AI Perimeter is at pre-launch. The founding structure below is the shape we are hiring into; named appointments will be published here as they are made, with prior affiliations disclosed.
| Role | Remit |
|---|---|
| Editor-in-chief | Editorial standards, corrections, final call on publication. |
| Agents & AI security reporter | Model and agent security, evaluation infrastructure, MLSecOps, vendor disclosures. |
| Threats reporter | Breaches, ransomware, nation-state activity, critical infrastructure. |
| Policy correspondent (Brussels) | EU AI Act enforcement, NIS2, CRA, and the divergence between EU, US, UK and APAC regimes. |
| Subscriptions and operations | Billing, subscriber support and the Intelligence dashboard. No editorial access and no say in coverage. |
Contact
News desk and tips: desk@aiperimeter.news. Corrections: corrections@aiperimeter.news. Subscriptions and billing: support@aiperimeter.news.
For sensitive material, we will publish a Signal number and a PGP key before launch. Until then, please do not send anything you would not want sitting in an ordinary inbox.