Editorial Standards and Corrections Policy

Mitzpe Institute publishes intelligence and analysis. The standard for that work is forensic sourcing, calibrated language about uncertainty, a clean line between contracted and public output, and visible corrections when a published claim turns out to be wrong. This page states the institute’s commitment to each of those, and the mechanism a reader can use when one of them appears to have failed.

For the separate standard a book must meet to carry the Mitzpe Press imprint, see Mitzpe Press’s Editorial Standard.

Sourcing

Every factual claim in a Mitzpe publication is verifiable. Primary sources take precedence over aggregators. Where sources disagree, the disagreement is named rather than resolved in silence. Where a claim rests on inference rather than reporting, the inference is shown.

The institute publishes in English. Source material is drawn from Hebrew, Arabic, and English-language reporting, and from the regional press in other languages where relevant. Translation of non-English source material is part of the production process. Where a translation choice meaningfully shapes the reading, the choice is flagged.

Where a quote from a named actor is load-bearing for an assessment — where the analytical claim turns on what the actor said and how it was said — the translation is checked against the original. Quotes that are illustrative rather than load-bearing rest on the production-pipeline translation.

The institute does not republish other outlets’ reporting as its own. Where a published assessment depends on another outlet’s reporting, that outlet is named in the body of the analysis, not buried in a citation block.

Calibrated probability

Mitzpe writes about uncertain futures. The institute uses calibrated probability language — “likely,” “probable,” “low confidence,” numerical ranges where useful — rather than the rhetorical hedging common in commentary (“could,” “may,” “some say”). Calibration is a discipline. It produces statements that can be tracked, challenged, and updated.

When the institute commits to a probability, it commits in writing. Past calls are not quietly rewritten when the world goes a different way. The track record stays where it was published, and updated assessments link back to the prior call rather than supersede it silently.

Anonymous sources

The institute uses anonymous sources sparingly and only where attribution would expose the source to retaliation, professional harm, or risk to safety. Anonymous claims are corroborated against a second source or named context before being included. The reason for granting anonymity is stated in the analysis (“a senior official briefed on the deliberations,” “an analyst at a partner institution who declined to be named”). Pseudonyms are not used.

AI in production

Mitzpe uses large language models throughout its production pipeline. Nothing publishes until a human editor has read it, changed what needed changing, and approved it. Mitzpe does not publish AI-generated commentary as human commentary. The byline names the analyst who owns the call, and that analyst answers for every line under it.

Corrections

When a Mitzpe publication contains a factual error, the institute corrects it visibly. Corrections appear at the top or bottom of the piece, dated, naming what was wrong and what is now right. The original incorrect text is preserved with a strikethrough where the change is in the body of the analysis, or noted in a correction line where the change is to a published number, name, or date.

Silent corrections are not the institute’s practice. A reader who saw the original piece should be able to see what changed. A reader who arrives later should see that the piece has been corrected, and what.

The threshold for a correction is a factual error — a misstated number, a misattributed quote, a wrong date, a misidentified actor. Updated assessments where the underlying facts shifted are not corrections. They are updates, and they are handled in the next published assessment with a link back to the prior call.

To request a correction, email corrections@mitzpe.org with the URL of the piece, the specific claim in question, and the source for the corrected information. Mitzpe responds to correction requests within five business days.

The client–editorial firewall

Mitzpe’s public analytical work is not gated by, shaped by, or pre-cleared with any contracted client. Where a public assessment touches on a sector or actor in which the institute has contracted work, the relationship is disclosed in the analysis where it is material to the reader’s reading.

The full statement of the firewall — what contracted clients can and cannot do, how it applies to government engagements, and how conflicts of interest are disclosed — lives on the Disclosures page. The editorial point on this page is the one that matters to a reader of Mitzpe analysis: nothing the reader sees in a Mitzpe publication has been edited, vetoed, or pre-cleared by a paying client.

Bylines and accountability

Every Mitzpe publication carries a byline. The byline names the analyst who owns the call. Where multiple analysts contributed to a piece, all are named. The institute does not publish under a house byline alone — a human analyst is identified for every piece.

Editors are also named where their structural intervention shaped the analysis. The reader knows who wrote, and who shaped what they wrote.

Voice on uncertainty

Mitzpe writes for readers who use analysis to decide. That register requires saying what is known, saying what is not, and not papering the gap between the two with rhetorical confidence the evidence does not support. The institute treats it as an editorial failure to overstate certainty — and an equal failure to understate it. A signal at low confidence is still a signal. A signal at high confidence carries a different weight, and the reader is told which is which.

Contact

To request a correction

corrections@mitzpe.org

For other editorial questions

shalom@mitzpe.org