DeepSeek: Critical Implications for Government Work

Over the weekend, a groundbreaking development in artificial intelligence made global headlines: the unveiling of DeepSeek, China’s latest AI language model. This innovation represents a serious contender in the AI landscape, demonstrating performance levels comparable to ChatGPT-o1. The accomplishments behind DeepSeek are nothing short of remarkable:

• Development Speed: Created in just two months, showcasing a rapid innovation cycle.

• Cost Efficiency: Training expenses were slashed by 94%, dropping from $100M to an impressive $6M.

While these achievements are undeniably a major leap forward in AI technology, they also present significant considerations for government employees. As Chair of DPIRD’s AI Working Group, I want to emphasise the critical importance of protecting organisational data and ensuring compliance with privacy policies. It is essential to approach tools like DeepSeek with caution and avoid their use for work-related purposes.

Key Considerations for Government Employees

1. Privacy Concerns: DeepSeek stores all user input on servers located in mainland China. This data is used to train the model, which directly contravenes our offshore data policy. Sharing sensitive or organisational information with such platforms could expose our operations to significant security risks.

2. Open-Source Availability: DeepSeek offers a local deployment option, enabling users to download and run the model on personal devices. While this avoids server-based privacy issues and ensures the model cannot train on your data, it still requires careful consideration.

Responsible Innovation in AI

DeepSeek’s technological advancements are a testament to the rapid pace of AI innovation. However, these developments highlight the necessity of maintaining rigorous data protection standards. As government employees, we must prioritise:

• Adhering to Organisational Guidelines: Avoid using external AI tools that do not comply with data security policies.

• Promoting Responsible AI Use: Through ongoing training and education, DPIRD aims to empower staff with the knowledge needed to navigate emerging AI technologies safely and effectively.

Instead of simply blocking access to platforms like DeepSeek, our goal at DPIRD is to foster a culture of responsible AI use that aligns with our organisation’s values and privacy standards.