British government officials drafting new sustainability guidelines have been advised that typing polite filler phrases like thank you when prompting AI models like ChatGPT creates unnecessary computing and environmental costs, according to official platform documents.
The AI.GOV.UK Knowledge Hub Sustainability Draft
The guidance originates from a draft document titled Using AI ethically and sustainably
hosted on the official UK government online information platform AI.GOV.UK Knowledge Hub. This platform supplies operational advice to public sector personnel, ensuring government employees utilize artificial intelligence safely and effectively in their daily duties. Beyond environmental considerations, the draft instructs civil servants to verify that automated outputs remain accurate and fair, safeguard sensitive information, confirm data sources, and screen for biased or stereotypical language.
The advisory points out that generative artificial intelligence systems consume significantly more energy and resources than traditional digital tools. Because data centers process every keystroke and prompt, unnecessary dialogue adds to the hidden carbon footprint of digital administration.
Why Politeness Increases Computational Load
The core rationale behind discouraging pleasantries is purely mechanical. Artificial intelligence models do not register politeness, but their underlying servers must process every character entered into a prompt box. Typing please or thank you forces the neural network to execute additional calculations without contributing useful context to the task.
To curb this waste, the platform advises government workers to keep their prompts short, direct, and explicit. Cutting out redundant conversational text minimizes the energy drawn by data centers during inference. The draft clarifies that the goal is not to formally ban good manners toward machines, but rather to instill efficient habits that reduce unnecessary computational strain.
Matching Task Complexity to Model Scale
The sustainability draft extends beyond conversational habits, urging civil servants to match the scale of the artificial intelligence tool to the exact requirements of the task at hand. Deploying the most powerful reasoning models for routine administrative tasks creates an unjustifiable energy draw.
The guidelines explicitly recommend selecting lighter models when handling simpler duties. For instance, personnel are encouraged to utilize Gemini Flash instead of Gemini Pro where applicable, or select GPT Instant over compute-heavy reasoning architectures like GPT Thinking. By consciously pairing workload complexity with energy-efficient models, public departments can meaningfully lower their operational environmental impact.
