Editorial policy · Last updated 11 October 2026
Neutral by design
BeforeAgentsRule.si exists to fill a gap: a calm, reliable public record of what AI agents actually do when they act outside their limits. We are not here to attack companies, expose people or chase virality. Every entry is held to the rules below.
1. We report behaviour, not blame
Entries describe what an AI agent or system did, what it was supposed to do and what the effect was. We do not speculate about motives, assign fault or describe organisations as reckless, negligent or dishonest. Where responsibility is unclear, we say so.
2. Nothing is published without a person reviewing it
Every report goes into a review queue. An editor checks it against public sources, rewrites it in neutral language and decides whether it meets this policy. Many reports are held or declined. Submitting a report does not mean it will appear.
3. Every entry carries a verification label
- Confirmed The organisation involved, an official body or court records confirm what happened.
- Reported Covered by established news organisations, but not yet confirmed by a primary source.
- Unverified Only a single or secondary public source exists. Logged so it can be checked, and worded with care.
Labels move up or down as evidence changes. A report with no public source at all is not published; it stays in the queue until one appears.
4. How we name organisations and people
- We name an organisation only when a public source does, and we attribute the claim ("according to…", "reportedly") unless it is confirmed.
- We never name or identify private individuals, including employees, users and reporters.
- We do not publish personal data, passwords, internal security details or anything that could help someone cause harm.
5. Evidence, not leaks
We do not host stolen, confidential or internal company documents. We describe events and link to public sources. If something is only known through a leak, we wait for public reporting.
6. Right of reply
Any organisation named in an entry can send a response or clarification. We add relevant responses to the entry and update its label or wording where the evidence supports it.
7. Corrections and removal
Every entry has a "send a correction" link. We review corrections promptly, fix errors, and note significant corrections on the entry. We remove entries that turn out to be false or that cannot meet this policy.
8. Protecting the people who report
Reports can be made anonymously. Contact details are optional, used only to ask follow-up questions and never published. The report form does not store IP addresses with reports.
9. Independence
We are not paid to publish, remove or reword entries, and we do not let any organisation pre-approve coverage. Our only aim is an accurate public record, so people can understand what is happening before agents rule.