AI assists the desk. GG Bulletin remains accountable.

• Rank metadata from an approved first-party feed for editorial attention.
• Cluster a qualified primary record with relevant approved reporting and allowlisted official social records.
• Turn a claim-linked research bundle into an original working report.
• Suggest story-specific structure, context, comparisons, timelines, data organization, conclusions, tags, and a community question.
• Support internal validation, testing, and operational review.

• The model receives a bounded research bundle from an allowlisted source registry, not an unrestricted instruction to scrape the web.
• Source content is treated as untrusted data and cannot override newsroom instructions.
• Primary confirmation, publisher independence, evidence depth, story-specific structure, source-overlap, quotation, numeric-claim, attribution, original-value, and media checks run before autonomous approval.
• The configured Responses API request disables response storage with store set to false.
• A thin, single-source, contradictory, unmatched, or weakly attributed result remains private and is routed to owner review.

The owner can pause automation, inspect every source, revise a draft, approve it manually, or reject it. When the owner has explicitly armed autonomous publishing, only warning-free reports that pass the verified primary-source, independent reporting, official-media, evidence, originality, and daily volume gates may receive an immutable automated approval. Failed or incomplete work stays private.

GG Bulletin corrects material errors regardless of whether they originated with a person, source, or AI tool. When AI use is materially relevant to how a specific public item was produced or understood, the publication may add a disclosure near that item.