How PrismaNews uses artificial intelligence.

Why you see the label

At the end of each briefing, we show a concise note describing it as an AI-assisted summary. To avoid repetition, the note is not repeated in list previews. Generated editorial images carry the compact “AI” badge.

The note and badge are not judgments about content quality; they make the role of automated systems in its production visible.

From sources to briefing

PrismaNews collects articles from public sources, normalises them, and groups those describing the same event. Generative models produce the title, standfirst, and briefing body using the source group available for that generation.

Sources linked to the briefing remain available. The divergence indicator compares how outlets and information regions describe the same event.

Automated checks

The PrismaNews Watchdog compares the briefing with the source snapshot used to generate it. It looks for unsupported claims, contradictions, material omissions, and non-publishable responses.

The check can pass, block, or request another generation. A positive result reduces error risk but does not guarantee perfect content.

Automated checks and human review

Watchdog is an automated control and does not amount to substantive human editorial review.

When a qualified person performs and records an editorial review, that status is stated separately. Without that statement, the briefing should be treated as generated and checked automatically.

Images, audio, and social content

Editorial images may be generated with AI and carry the compact “AI” badge. Audio and social content may be produced from the same briefing using speech synthesis or generative systems.

Synthetic content that could be mistaken for an authentic recording requires a visible or audible warning before publication.

Limitations, corrections, and version

Generative systems can introduce errors, inaccurate attribution, or incomplete context. Always inspect the linked sources when a decision depends on content accuracy.

To report an error, email [email protected] and identify the relevant briefing.

Transparency policy AI-T1 · effective 2 August 2026.