Recognising the urgency of AI sustainability

• Participants acknowledged the rapid expansion of generative AI across digital media, alongside increasing pressure from regulators, customers and internal stakeholders to understand its environmental implications.

Energy and emissions as the starting point

• There was broad agreement to prioritise energy use and carbon emissions, with water and other impacts recognised as important but less tractable in the near term.

Transparency gaps remain significant

• Limited disclosure from closed‑source models and upstream providers was identified as a key barrier, reinforcing the need for standardised frameworks and clearer expectations around reporting.

Enterprise leverage through procurement

• The group discussed how customer demand, procurement language and contractual requirements can be powerful levers for driving better disclosure and more sustainable AI practices.

Three layers of AI use clarified

• Participants distinguished between internal productivity tools, customer‑facing AI features, and model training - helping to focus effort on where most organisations have influence and exposure.

Context matters as much as numbers

• There was strong consensus that emissions figures are only meaningful when paired with clear boundaries, assumptions and comparison points, particularly given fast‑evolving technology.

Managing polarised narratives

• The group noted the highly charged public discourse around AI, agreeing that DIMPACT can play a valuable role in providing balanced, evidence‑based insight that avoids extremes.

Foundation for future research

• Participants supported developing a shared framework, collating existing knowledge, and identifying key research questions rather than chasing precise figures prematurely.

Momentum agreed

• To maintain engagement, the group agreed to meet monthly initially and begin scoping a paper to organise, synthesise and share insights relevant to digital media organisations.

The session established a strong foundation for collaborative, evidence‑led work on AI sustainability.