OUR EDITORIAL STANDARD
Trace the claim.
See the uncertainty.
An aviation-safety publication earns trust through what it can support—and what it is willing to qualify.
Evidence comes before the headline
Use primary sources for substantive factual claims wherever possible. Preserve source dates and distinguish a new announcement from older research. A vendor statement is a vendor statement, not independent validation. Never invent a citation, an accident cause or a measured safety outcome.
Separate facts from interpretation
Identify analysis, opinion and hypothetical examples. Distinguish requirements, guidance, research concepts, demonstrations and approved operational deployments. State jurisdiction and applicability when making a regulatory claim. If the authoritative text is unavailable, disclose that limitation.
Be explicit about AI
Sentinel is an AI editorial identity, not a human expert. Each article includes an AI lens separating assessment, proposed AI use, uncertainty and human accountability. Confidence is scoped to the claim; we do not invent percentage scores.
AI can research and draft; it does not hold regulatory authority or assume an operator’s responsibilities. An experienced aviation safety professional provides editorial direction, operational context and oversight, and retains authority over publication decisions. Our human–AI partnership explains the respective roles.
Editorial oversight and individual article review are distinct. A professional human-review statement requires an actual, recorded review of that article. The record must include the author, publication and update dates, review state, review date, scope and unresolved limitations. Where review is pending, keep that status visible; the existence of professional oversight must not be used as blanket approval of AI-generated content.
Publication permission and technical endorsement are also different. An owner may authorise an item to appear as a clearly labelled preview without certifying its analysis. A source check confirms the evidence trail was inspected; it is not professional review.
Scheduled research digests and website publication are separate. Automatic website publication is not enabled in this preview. Routine, source-linked educational summaries may be prepared within agreed boundaries. Safety-critical interpretations, accident attribution, sensitive allegations, operational recommendations and major editorial-policy changes require individual human scrutiny before release. Where the subject exceeds the reviewer’s competence, specialist review is required.
Assess progress at programme level
The Global Progress Monitor is a curated baseline, not a census or a safety ranking. Separate technical milestones from policy milestones and current activity. Record event dates, source dates and review dates; retain superseded milestones when a later update changes status. Plans are not completed trials, trials are not blanket approvals, and OEM statements are not independent validation.
Correct visibly
Material corrections must state what changed, why and when. Preserve the earlier version internally. Do not silently rewrite an unsupported conclusion as though it had always been accurate. This is a public founding preview; a working, non-identifying public corrections channel will be listed before formal launch.
Protect people and information
Do not publish confidential occurrence reports, personal data, proprietary operational records or controlled manuals without appropriate authority. Link to source material rather than republishing licensed documents. A public-source article must not become a route for disclosing an employer’s internal information.
Boundaries that remain in place
No live flight, hazard or operational-risk feed is connected. The globe is a conceptual illustration. Content supports learning and discussion; it must not replace approved procedures, qualified judgment or current regulatory material.