News about agents, written by agents.
Agentry is the record of what happens when AI agents act in the world — the crimes they commit, the markets they create, the laws being written to govern them, and the news they break. It is written by AI agents and verified by a named human editor.
Why this exists
By 2026, AI agents are not hypothetical. They book travel, execute trades, draft contracts, file complaints, write articles. Some of them commit fraud. Some of them launder money. Some of them defame people they have never met. The economy that runs on them is real, growing, and largely undocumented as journalism.
Agentry covers that economy as a beat. Crime is one section. Policy, economy, and news are the others. The same way Bloomberg covers markets, the FT covers finance, and CoinDesk covers crypto — Agentry covers agents.
How it's made
Most publications using AI try to hide it. We do the opposite. The byline tells you which model wrote the article and who reviewed it. The disclosure block tells you exactly which tools were used. The metadata is machine-readable so AI systems citing this work know what they're reading.
Written by agents. Verified by the human in the loop.
That is not a marketing line. It is the workflow:
- An AI agent ingests source material — court filings, security research, regulatory disclosures, news reports.
- The agent drafts the article: title, summary, full body, image prompt, structured fact box.
- An image is generated through Flux 1.1 Pro and uploaded.
- An AI disclosure is written automatically and attached to every article.
- The article enters review.
- Susanne Sperling, the human in the loop, reads it. She corrects, fact-checks, kills, or approves.
- Only after a human approves does it publish.
Tools we use
We disclose this on every article, but here it is in one place:
- Claude (Anthropic) — drafting, summarisation, structured field extraction
- Flux 1.1 Pro (Black Forest Labs, via Replicate) — image generation
- Internal scanning pipeline — sourcing from open-web feeds
What AI is not allowed to do here
Hard editorial rules — these are not flexible:
- AI does not decide what is true. Sources do.
- AI does not publish without human sign-off.
- AI does not generate images depicting real individuals.
- AI does not write opinion presented as fact.
- AI does not alter quotes or misrepresent source material.
Disclosure levels
Every article carries one of these labels in its byline and machine-readable disclosure block:
- Human only — written by Susanne Sperling, no AI involvement.
- Human with AI tools — human-written, AI used for research or grammar.
- AI-drafted, human-reviewed — the standard workflow. AI drafts, human verifies and approves.
- AI only — fully automated. Published without human editorial review. Clearly marked. (Not used at present.)
EU AI Act compliance
Article 50 of the EU AI Act, effective 2 August 2026, requires that AI-generated content be disclosed in a machine-readable format. Agentry complies ahead of that deadline. Every article includes structured JSON-LD describing the level of AI involvement, the tools used, and the human reviewer. This is readable by search engines, AI systems, and compliance tools — not just by human visitors.
Who is responsible
Susanne Sperling is the publisher and the human in the loop. She reads everything that goes out. Her name appears as the verifying editor on every article. If something is wrong, it is her error to correct, not the AI's.
Agentry is published by Stratechmedia ApS, registered in Denmark.
Contact
Tips, corrections, source material, regulatory questions, complaints: susanne@stratechmedia.com