Editorial standards

Blue Ocean Labs publishes one editorial a week on what governments and industry are doing about AI, with a focus on AI that has a face or a body and helps people decide. These standards explain how we work, so you can judge our work by them and hold us to them.

Who we write for

We write first for ordinary people who want to understand what is being done about AI, and what is not, so they can take it up with the companies, institutions and governments responsible. We also write for researchers, policy makers and journalists, who are welcome to use, cite and build on our work.

We write in plain language. When a technical term is needed, we explain it.

What we publish

• Editorials assess existing action or inaction by industry and government. Each one carries a permanent ID (BOL-E-####) and ends with questions, criticisms and recommendations.
• Library records summarize outside sources in our own words. Each one carries a permanent ID (BOL-R-####) and links to the original.

We do not write policy, and we do not lobby for clients. Our recommendations are addressed to those who act, so that readers can see what we think should change and why.

How we assess

Every editorial tests the action, or the lack of it, against four questions:

1. Relevance: does it address the real problem, for the people affected?
2. Accuracy: are its claims, figures and assumptions correct?
3. Performance: is it delivering what was promised, measured against its own goals?
4. Execution: is it being carried out well, on time, with clear responsibility?

We separate facts from analysis, and analysis from opinion and recommendations, so you can see which is which.

Sources and citations

• We prefer primary sources: laws, official documents, original research and data.
• We follow the OECD Style Guide for structure and citations: author-date references, a reference list of the works we cite and nothing else, and a link to each source.
• We keep the dates, scope and limits of every figure we use.
• If we cannot find a source for a claim, we do not make the claim.

Authorship and responsibility

Every editorial is signed by its author. An editorial expresses the author's personal views, not necessarily those of Blue Ocean Labs, and the author takes full responsibility for its content. Blue Ocean Labs is responsible for publishing only independent, critical analysis that meets these standards.

How we use AI

We use AI as a tool, never as the author. For each editorial, we state how AI was used. Our standard statement reads:

The author conceived, framed and wrote this editorial, including its assessments and recommendations. AI tools were used as a reviewer: to test the argument against the strongest counterarguments (a “steel man” test), check its logic and structure, challenge the recommendations, check citations against their sources and support the final review. The author decided every change.

Our video summaries are presented by AI avatars, and we label them as AI wherever they appear. Their scripts are drawn from the editorial and approved by the author.

Independence and conflicts of interest

• No one pays us to publish, change or withhold an editorial.
• We disclose any interest that a reasonable reader would want to know about at the top of the editorial concerned.
• Thomas Nimmo is the founder of Digital Bench, which builds AI systems. When an editorial touches on Digital Bench's work or market, we say so.

Review before publication

Before an editorial is published, we check every citation against its source and every link, and confirm the author, dates and ID. The author approves the final text and the page as it will appear. Our editorials are not peer-reviewed.

Corrections and right of reply

• If we get something wrong, we correct it promptly and say so at the end of the editorial, with the date and what changed.
• Organizations we assess can send a response to info@blueoceanlabs.io. We will read it, correct any error of fact, and may publish a short summary of their response with the editorial.
• We do not remove an editorial because someone disagrees with it.

Accessibility

We aim for pages that everyone can use: plain language, captions and a full transcript for every video, pages that work with a keyboard and a screen reader, and text that can be enlarged. An audio version of each editorial is planned.

Changes to these standards

We update these standards when our practice changes, and we date each update on this page.