Shaping the AI sustainability narrative

• The group discussed how publisher‑specific AI use cases - such as discovery tools, knowledge management, translation and editorial support - differ from broadcaster and streaming use cases, and the importance of ensuring these are meaningfully represented in forthcoming AI research.

Focusing on relevant AI use cases

• Rather than headline content generation, participants highlighted embedded and operational AI uses within publishing products and workflows as priority areas for deeper analysis.

Counterfactuals and real impact

• Significant discussion centred on how to assess AI’s environmental impact relative to what it replaces, acknowledging that counterfactuals vary widely by use case and introduce challenging boundary and lifecycle questions.

Navigating uncertainty and rebound effects

• The group reflected on the tension between theoretically robust consequential analysis and what is practical, communicable and useful for organisational decision‑making and Scope 3 targets.

Indirect and downstream impacts

• Participants raised challenges around limited transparency when publisher content is used by third‑party AI systems, making indirect use‑phase emissions particularly difficult to estimate.

Supplier data realities

• Feedback on the supplier data request form highlighted mixed experiences between procurement and sustainability teams, with strong interest in the tool but recognition of supplier capacity limits, particularly for large cloud providers.

Product vs corporate footprints

• The session underscored the importance of clearly distinguishing between product carbon footprints and suppliers’ wider corporate emissions when asking for data.

Iterative, collaborative approach agreed

• The group supported refining the supplier data request process over time, including the idea of optional modules for different supplier types, and using the tool as a learning and engagement mechanism rather than a one‑off request.

The session reinforced the need for pragmatic, publisher‑led approaches to AI sustainability, grounded in real use cases, transparent about uncertainty, and focused on informing better design and procurement decisions as the landscape continues to evolve.