Key Takeaways
• Bhutan launches its first interim guidelines for the use of generative AI in government services
• Guidelines focus on ethical use, data privacy, human oversight, and avoidance of sensitive data inputs
• Developed using policy references from global leaders like Canada, the UK, the U.S., and the EU
• Civil servants are warned against using generative AI as search engines unless sources are verifiable
In an effort to embrace technological innovation responsibly, Bhutan’s Royal Civil Service Commission (RCSC), in collaboration with the GovTech Agency, has issued its first interim guidelines on the use of generative artificial intelligence (AI) within the public sector.
The directive marks a significant step toward structured digital transformation while maintaining a strong stance on ethical use, privacy safeguards, and national security.
Purpose and Framework of the Guidelines
The newly released document acknowledges the potential of generative AI to boost public sector efficiency, but also emphasizes the risks it poses if misused.
The policy provides interim direction in the absence of a national AI law and draws on international AI governance frameworks from countries such as Canada, the United Kingdom, Switzerland, the United States, the United Arab Emirates, and the European Union AI Act.
What the Guidelines Emphasize
• Ethical AI usage: Ensure alignment with national values, ethical standards, and inclusivity
• Human oversight: AI must assist—not replace—decision-making, particularly in sensitive operations
• Verifiable sources: Civil servants must not use AI tools as search engines unless they cite trusted sources
A core component of the framework is its focus on educating civil servants about AI’s risks. They are encouraged to understand issues around bias, diversity, inclusion, anti-racism, and ethical AI behavior to better scrutinize algorithm-generated content.
Furthermore, they must label all AI-generated outputs clearly and review content for contextual and factual accuracy prior to use in any official capacity.
Data Privacy and Content Sensitivity Protections
The guidelines raise serious concerns about data collection practices embedded in many AI platforms. Tools like ChatGPT and Gemini collect user inputs and metadata unless specifically opted out of.
To protect personal and institutional data, Bhutan’s civil servants are instructed to:
• Avoid entering personally identifiable information, including names, ID numbers, and email addresses
• Never input unpublished materials or any government data classified as Level Two, Three, or Four
• Use de-identification methods, such as substituting sensitive content with format-consistent placeholders
Officials are also warned that inputting such data could result in exposure to third-party platforms, compromising state confidentiality and intellectual property.
AI Must Not Replace Human Judgment
Perhaps the most crucial element of the document is its reminder that generative AI is not a replacement for human reasoning in public service. AI can aid in areas like drafting reports or generating templates but cannot be the final authority in critical matters.
This requirement reflects global best practices, where accountability in areas like hiring, budgeting, and education must remain firmly under human control.
Although Bhutan is in the early stages of AI policy development, this interim guideline offers a measured, globally-aligned foundation for future legislation.
By adopting a proactive stance, Bhutan positions itself among nations trying to balance innovation with responsibility, without waiting for reactive measures after potential harms surface.
It also serves as a model for other emerging economies, highlighting that even without a fully established AI infrastructure, governments can act decisively to protect their institutions and citizens in the digital age.
January 23, 2025: Musk Criticizes Trump’s $500B AI Plan, Claims Backers Lack Funds! April 11, 2025: Mira Murati Launches $2B AI Lab ‘Thinking Machines’ to Redefine Artificial Intelligence April 8, 2025: Ripple’s $1.25B Deal Sparks AI Finance Shake-Up
For more news and insights, visit AI News on our website.
