This article is the third in a series discussing AI tools, with a focus on Claude Cowork. The first entry emphasized the importance of taking time and not rushing decisions. The second explored AI and privacy concerns to be aware of. This piece outlines the framework the author uses for deciding on AI usage, advising caution since decisions impact more than just personal risk. The framework consists of three key considerations: employer, personal privacy, and others’ information. This approach helped determine what to allow Claude access to during recent experiments with Claude Cowork. Although circumstances vary, the author hopes sharing their reasoning provides a helpful guide.<\/p>\n
For those employed by an organization, it\u2019s important to consider employer guidelines before using AI tools for work. Policies differ by location and in the U.S., they vary by state. Employers typically have rules about the use of company equipment and systems for work purposes. Familiarity with an employer’s AI policies is crucial, especially regarding data protection in specific fields, like FERPA in higher education. The author uses personal devices and Microsoft tools, which comply with institutional privacy measures, for work-related tasks. Claude Cowork is reserved for non-university-related activities. It is advisable to assess if your institution has an AI policy, what data you might share, and which AI tools are approved for use.<\/p>\n
Personal privacy comfort levels demand introspection, as they are highly subjective. Many haven\u2019t fully considered privacy implications, particularly in the context of AI. Resources from Civics of Technology can aid in understanding privacy better. Personal data, like journals and health information, is kept away from AI tools like Claude. The author uses Day One for journaling, valuing its encryption. Publicly available information, like podcast content, is less restricted in AI-accessible spaces. The author advises configuring AI tool settings regarding data training permissions and assessing app contents individually to protect sensitive information.<\/p>\n
Protecting others’ information is a priority, requiring careful consideration. The author does not permit AI access to contacts or personal identifiers of others. Publicly available details, such as those from the Teaching in Higher Ed podcast, are managed differently. Information like episode details and workshop notes, which are publicly shared, may be stored in Claude-accessible notes. However, private or confidential discussions remain inaccessible to AI. The approach is to limit AI access to information already available on the web.<\/p>\n