TLDR: Delaware’s Department of Technology and Information has enacted a new policy governing the use of generative AI tools by state employees. The policy, effective July 3, 2025, allows the use of public AI tools like ChatGPT and Gemini, as well as enterprise versions such as Microsoft Copilot and Amazon Bedrock. However, it explicitly bans ‘malicious AI models’ like DeepSeek and RedNote, both developed by Chinese companies, citing concerns about data security and potential misuse. The policy emphasizes responsible AI use, data protection, and encourages mandatory training for employees.
Delaware’s Department of Technology and Information (DTI) has officially adopted a comprehensive policy outlining the permissible use of generative artificial intelligence (GenAI) tools by state employees. The policy, dated July 3, 2025, aims to provide a clear framework for state functions and data sharing while mitigating potential risks associated with AI adoption.
The new guidelines permit state employees to utilize a range of public and enterprise AI models. Approved public tools include popular platforms such as ChatGPT and Gemini. For enterprise-level applications, the policy greenlights solutions like Microsoft Copilot and Amazon Bedrock, which are designed with enhanced data protection and integration capabilities for state systems.
Crucially, the policy explicitly prohibits the use of ‘malicious AI models,’ specifically naming DeepSeek and RedNote. Both of these models are developed by Chinese companies, and their prohibition underscores concerns regarding data security and potential vulnerabilities. The DTI’s stance reflects a cautious approach to AI tools that may pose risks to sensitive state data.
A cornerstone of the new policy is the stringent prohibition against uploading any sensitive state-owned data into public AI tools. This measure is designed to prevent such data from becoming publicly accessible or being inadvertently used to train AI models, thereby safeguarding confidential information.
Furthermore, the policy mandates that employees independently verify all AI-generated output to prevent the dissemination of false, biased, or harmful information. It also restricts the use of AI tools to official state purposes only and prohibits their use on personal devices or accounts to circumvent established policies and safeguards. Employees are also urged to clearly label any AI-generated content in both internal and external communications.
Also Read:
- Concerns Mount Over DeepSeek AI’s Bias, Jeopardizing China’s Global AI Ambitions
- Federal Agencies See Ninefold Surge in Generative AI Adoption, GAO Reports
To ensure responsible implementation and mitigate cybersecurity risks, the DTI strongly encourages agencies to require employees to complete specific training before granting them permission to use generative AI tools for work-related tasks. Anthony Collins, DTI director of enterprise architecture and solution integration, highlighted that the policy serves as a crucial first step in guiding employees on acceptable AI use and protecting valuable state data assets. The Delaware AI Commission is also developing employee training related to responsible and principled AI use, and is exploring a ‘sandbox’ for testing innovative AI technologies.
