*Environmental focus: include focus on environmental impacts when considering using gen AI solutions, compare to alternative tech that do not use LLMs — check provider renewable energy, $AMP $RARE $2Z energy-efficient infra, select low-carbon grids*
- *Smallest capable model: "When selecting models, choose the smallest (in terms of number of parameters) that meets your requirements as these are likely to have the lowest environmental impact" — your Gemini Flash over Pro / GPT Instant over Thinking example is correct interpretation, though framework doesn't name commercial models*
- *Other mandatory principles in same doc: never put sensitive/personal/classified/intent of government data in — output susceptible bias/misinformation, must be checked thoroughly and cited appropriately — lifecycle analysis + transparency on carbon cost + mitigation e.g. energy-efficient hardware*
- *Context Sep 2026: UK pushing sustainability labels + measurement, not new binding AI usage ban — other recent UK AI rules are: under-18 mandatory breaks for chatbots, Online Safety Act expansion to ChatGPT/Gemini, CMA Dec 2026 publisher opt-out for AI Overviews — environmental guidance remains framework level, not statutory force*
*BREAKING 🚨 UK HM Government Generative AI Framework reshared as green guidance 🚀 Use AI only when essential check spreadsheets/search first — compare env impact vs non-LLM — pick provider with renewable + low-carbon grid — choose smallest params model that meets need — lifecycle carbon audit + transparency — all AI output must be verified for bias/misinformation + disclosed 📢*
- *Smallest capable model: "When selecting models, choose the smallest (in terms of number of parameters) that meets your requirements as these are likely to have the lowest environmental impact" — your Gemini Flash over Pro / GPT Instant over Thinking example is correct interpretation, though framework doesn't name commercial models*
- *Other mandatory principles in same doc: never put sensitive/personal/classified/intent of government data in — output susceptible bias/misinformation, must be checked thoroughly and cited appropriately — lifecycle analysis + transparency on carbon cost + mitigation e.g. energy-efficient hardware*
- *Context Sep 2026: UK pushing sustainability labels + measurement, not new binding AI usage ban — other recent UK AI rules are: under-18 mandatory breaks for chatbots, Online Safety Act expansion to ChatGPT/Gemini, CMA Dec 2026 publisher opt-out for AI Overviews — environmental guidance remains framework level, not statutory force*
*BREAKING 🚨 UK HM Government Generative AI Framework reshared as green guidance 🚀 Use AI only when essential check spreadsheets/search first — compare env impact vs non-LLM — pick provider with renewable + low-carbon grid — choose smallest params model that meets need — lifecycle carbon audit + transparency — all AI output must be verified for bias/misinformation + disclosed 📢*
