How this portal is researched and written
The standards behind every analytical piece. Sources, framing, ethics guardrails, and what we will and will not cover.
Every piece on digital-humans.org begins with primary sources. For coverage of foundation models, that means vendor documentation, lab announcements, technical reports and peer-reviewed papers. For voice and avatar platforms, it means direct evaluation alongside vendor materials. For policy and trust-and-safety coverage, it means government documents, court filings, organisational publications and the academic literature. Secondary sources — trade publications, analyst reports, industry surveys — are used to triangulate, never as the sole basis for a claim.
Source hierarchy
The site treats four tiers of source with descending weight. Primary sources — vendor documentation, papers, official filings — carry the most weight and are linked directly. Peer-reviewed research carries comparable weight in technical claims. Established trade publications — The Information, MIT Technology Review, Stratechery, Nature, Bloomberg — are cited for industry context and corporate analysis. Aggregators and unverified claims are not used as the basis for assertions; where they appear, they are clearly marked as the kind of claim they are.
Living trackers
Two pieces on the site are explicitly living trackers — the open-source AI weekly synthesis and the AI startup funding monthly. These are date-stamped on every update, and previous versions remain accessible. They are analyst synthesis, not aggregation. Where a tracker piece references events from the past month, it does so to identify patterns rather than to list every release. The tracker pieces are the only articles on the site with a cadence; everything else is published when the analytical work is complete.
Ethics-heavy coverage
Several areas on the site carry a strict analytical-only rule. Deepfakes and synthetic non-consensual intimate imagery, voice cloning of identifiable individuals without consent, face-swap technology in the context of harassment or fraud, and AI-generated content depicting real political figures are covered analytically — the legal landscape, the detection ecosystem, the policy response, the documented harms — but never instructionally. Coverage of platforms that enable these harms is descriptive, never a recommendation. The autonomous weapons debate is covered as a debate, with credible positions on multiple sides; the same applies to existential-risk discourse and the labor-market effects of AI.
The Anthropic exception. Because digital-humans.org is an independent analyst portal, it covers Anthropic, Claude, and the rest of the Anthropic ecosystem with the same analytical lens it brings to OpenAI, Google DeepMind, Meta and the others. This is the only site in the broader publication network where Anthropic is on the same analytical footing as any other lab.
Citation format
Every claim that follows from a specific source is linked inline. Where multiple sources support a claim — the standard pattern for technical writing — sources appear at the end of the relevant section as a numbered footnote-style list. Article footers carry a cross-link block: previous pieces on adjacent topics, related silos, and the canonical pillar within the current silo.
What we will not do
We will not aggregate press releases. We will not write recommendation content that ranks products by stars or scores. We will not provide instructional content for harms. We will not accept guest posts, sponsored content, or syndication. We will not republish content from other publications under a new title. We will not maintain affiliate links. Coverage is independent and uncompensated.
Corrections
Errors are corrected with a visible correction note at the foot of the affected article. Substantive corrections — those that change the meaning of a claim — also update the article's modification date. Cosmetic corrections (typography, broken links) are made silently. Send corrections to editor@digital-humans.org.