Ethics

This page is the public version of how we make editorial decisions at Rhinoceros Media. The internal version is more granular and is updated as we learn. The principles below are the parts that readers, sources, subjects of coverage, and other journalists should be able to hold us to.

How we verify

The standard we apply to every factual claim is defensibility: a hostile attorney, opposing researcher, or skeptical journalist should be able to verify the claim using only the citation we provide. If they can't, the citation is insufficient and the claim does not run.

We rely on primary records wherever possible: government filings (IRS Form 990, IRS Form 8872, LARA corporate registry, federal and state court dockets), municipal records (council packets, commission minutes, FOIA returns), campaign-finance disclosures (Michigan Bureau of Elections, FEC), property and tax records (county assessor, register of deeds), and statute (Michigan Compiled Laws, federal codes). When we cite news coverage, we archive the page and cite the archive URL alongside the live one.

Aggregate statistics carry their methodology with them. When we write “$3M+ in political spending” the citation tells you which database, which date range, which search parameters, and how the total was computed.

Findings versus inferences

We separate verifiable facts from analytical conclusions in the text itself. A finding (“Council Member X chairs both the Planning Commission and the Zoning Board of Appeals”) is stated directly. An inference drawn from that finding (“this raises questions about appellate independence”) is flagged with explicit hedging language. We do not present inferences as findings.

Where a relationship is unconfirmed (a family connection inferred from a shared surname; a financial relationship inferred from circumstantial evidence), we say so in the same sentence: “Connection unconfirmed; based on shared surname and geographic overlap.”

Quotes from public meetings

A direct quote attributed to a named speaker at a public meeting (council, commission, board, or committee) is a record of what was said, not an assertion by us that the underlying claim is true. We attribute by name and timestamp or page citation so readers can verify the quote against the recording or transcript. We do not add a “we did not seek a response” caveat to public-meeting quotes; the attribution itself signals that the quote is testimony or advocacy, not our finding.

We apply this symmetrically. If one speaker's direct-experience account stands without hedging, another speaker's equivalent quote in the same post does too. Hedging one and not the other is itself an editorial choice and we don't make it.

Paraphrased speech (someone summarizing or repeating from memory what another person said) is flagged explicitly. We don't present a paraphrase as if it were the original speaker's direct quote, and we try to locate the original recording before publishing.

Subjects of coverage

When a post makes an allegation against a named person or organization, we seek a response and reflect it in the post. If the subject declines or doesn't respond by deadline, we say so and note when and how we contacted them. If we choose not to seek a response (for example, when the post is purely about public-meeting record), we say that explicitly.

We don't grant quote approval. We don't accept conditions on what can be reported about the same matter from other sources. We don't publish stories the subject has paid to review.

Confidential sources and tips

We accept confidential tips through two end-to-end encrypted channels: Proton Mail and Signal. The tip line is staffed by editorial. We do not share the contents of tips outside that channel except to verify a claim, and we do not share the identity of a confidential source with anyone inside or outside the publication unless the source authorizes it.

Information from confidential sources is corroborated against primary records before it appears in a story. We do not publish claims based solely on a single anonymous source; the role of a confidential source is to point us at where to look, not to be the citation.

The tip-line device runs disappearing messages by default, screen-locked with a strong PIN, registration-locked against remote takeover, and not backed up to any cloud. Operational details we choose not to publish are kept that way deliberately.

AI and automation

We use language models in production work in a narrow, well-disclosed way: pattern-matching across large public records (council packets, FOIA returns, campaign-finance batches) to surface candidates for human review. Every fact that appears in a published post is verified against its primary source by an editor before publication. We do not generate prose, headlines, or quotes via language models. We don't treat any automation as a substitute for the defensibility standard above.

When automation contributes to a finding, we say so. If a spreadsheet was machine-generated and reviewed manually, the post notes that. If a claim was first surfaced by an automated scan and then verified, we cite the verifying primary source, not the scan.

Conflicts of interest

We disclose every commercial relationship we have with any entity that is, has been, or could plausibly become a subject of our coverage. The current conflicts register is below.

• None at this time. Rhinoceros Media has no advertising contracts, sponsored content arrangements, paid partnerships, consulting engagements, or commercial relationships with any entity covered in our reporting.

This list is updated whenever a relationship changes. Past versions of this page are visible in the site's git history. If we add a relationship that could affect future coverage, we add it here on the same day and include it as a recusal note in any post that touches the subject.

Editors and contributors disclose financial and personal relationships internally before participating in any coverage that could touch them. Recusal is the default when a conflict is identified.

Funding and editorial independence

We are funded by reader subscriptions, reader-sponsored FOIA contributions, and (when we open them) general donations. Subscribers and donors do not see stories before publication. They do not influence what we cover, what we publish, or how we frame it. Records that come back from a sponsored FOIA are published here on the FOIA log regardless of who paid the fee.

We do not accept advertising. We do not accept sponsored content. We do not accept payment, gifts, travel, or hosted access from any entity that is, has been, or could plausibly become a subject of our coverage.

No investigation is ever placed behind a paywall, in any format. Paid memberships buy extras (process content, physical goods, events, early access), never access to the primary work.

Corrections and updates

We distinguish between two things:

• Corrections change a factual claim because the original was wrong. Every correction is logged at /corrections, marked at the top of the affected post, and propagated to every other format the original claim appeared in (newsletter, social posts, zines, public installations).
• Updatesadd new substantive information after publication without contradicting the original. They are marked at the top of the post with the update date and a brief note. Updates are not the same as corrections, and we don't silently treat one as the other.

We don't silently edit posts to fix errors. We don't remove published posts to avoid embarrassment. If a post should not be public anymore (for example, a source became identifiable after publication), we explain why on the URL where the post used to be.

Embargoes and pre-publication review

We don't accept information under embargo from government, corporate, or political-campaign sources. Every fact we publish should be available to the public when we publish it.

We don't share drafts with subjects of coverage before publication. We do contact subjects for response on substantive allegations, with the specific allegation and a deadline, and reflect their answer (or lack of answer) in the published post.

Right of reply

Anyone named in a post can write to us with a response, a correction request, or a different reading of the record. We read every such message. If a correction is warranted, we issue one promptly. If the response is substantive but not a correction, we may quote it in an update or in subsequent coverage. We don't maintain a separate “letters to the editor” channel; the same contact channel is the place.

What we do not do

• We don't accept advertising, sponsored content, or paid placement.
• We don't grant quote approval, headline approval, or pre-publication review to subjects of coverage.
• We don't treat off-the-record as the default for government-source conversations. The default is on-the-record; sources can request off-the-record but they need to ask.
• We don't generate posts, headlines, summaries, or quotes with language models.
• We don't write SEO content, listicles, or aggregation posts of other publications' reporting.
• We don't pay sources for tips or interviews.
• We don't share readers' or members' data with any third party. The newsletter list is never sold, rented, licensed, or shared.

Found a problem with our work?

If you spot a factual error, a missing disclosure, or an editorial decision you disagree with, write to us via the contact page. We answer every message that requires an answer, and we publish corrections in public on the day we issue them.