AI news ·
Scottish Government Warns Officials: Only Share AI Inputs You’d Share Publicly
The Scottish Government advises officials not to share sensitive or unpublished information in AI tools like ChatGPT to avoid risks. Always fact-check AI outputs to prevent misinformation.
Scottish Government Issues Clear Guidance on AI Use for Officials
Officials working for the Scottish Government have received explicit instructions on how to handle information when using AI tools such as ChatGPT, Google Gemini, DeepSeek, and X’s Grok. The guidance stresses that civil servants must avoid sharing any government policy details or sensitive data that wouldn’t be suitable for public disclosure.
The key takeaway is simple: do not input anything into AI systems that you wouldn’t share openly with a member of the public. This includes embargoed announcements, campaign materials, and any details related to ministerial announcements.
Why This Matters
Generative AI tools operate like vast databases. When sensitive or speculative queries are submitted—for example, hypothetical government policies—they can inadvertently associate the government with information that may later appear in other users’ searches. This creates a risk of linking official institutions to content that is inappropriate or misleading.
The guidance provides a clear example: asking “what might happen if the Scottish Government set the driving age to 12?” could cause the government’s name to be connected with this idea in AI-generated text. Such associations can lead to unintended reputational issues.
Practical Rules for Civil Servants
- Do not share sensitive or embargoed information in AI tools.
- Avoid discussing material related to campaigns or ministerial plans.
- Consider whether paid AI services have sufficient safeguards for the data you submit.
- Always fact-check AI-generated information. AI can produce “hallucinations”—responses that seem factual but are false or misleading.
This guidance follows recent reports of Scottish councils using AI to draft reports and communicate with constituents in a clear, concise way. It also aligns with policy shifts at the UK Department for Work and Pensions, which recently relaxed its ban on public AI tools, except for DeepSeek which remains prohibited.
Ensuring Responsible AI Use in Government
For government employees looking to better understand AI’s capabilities and risks, training resources can be valuable. Platforms like Complete AI Training provide courses on safe and effective AI use tailored for professionals in government roles.
Following this guidance will help maintain the integrity of government communications and prevent unintended leaks or associations that could undermine public trust.