AI Prompt Management
21. Why is BW monitoring all AI prompts?
Including confidential data in an AI prompt risks exposing sensitive information to the model's developers or third-party reviewers who monitor inputs for quality and safety. Furthermore, many AI platforms use user prompts to retrain their models, meaning your proprietary and perhaps very personal data could potentially resurface in responses generated for other users.
22. Is BW using new software for AI prompt monitoring?
Yes, BW is using SentinelOne’s Prompt AI. BW has used SentinelOne software for several years now to protect our PCs and servers. Using SentinelOne’s Prompt AI will help ensure that BW’s IT security tools all work well together.
23. Can I use any AI after Share Smart goes live?
Once Share Smart goes live, you will be able to use most AIs. This is a big change from the current policy, where BW limits access to around 50 AI sites! Some AIs have been classified as unsafe, and access to those AI tools will be blocked. We will also block access to the same category of AI sites we block today for internet browsing.
The goal is to allow broad AI use while reducing information security risk.
24. Will everything I submit to an AI be monitored?
BW will inspect all AI prompts submitted in a web browser (except for those submitted using Microsoft M365 Copilot) in real time in order to look for policy violations (for example, confidential or personal data, as well as prompts about subjects that violate BW’s cultural values, such as pornography).
25. Does Prompt AI save my prompt content?
No! Prompt AI does not log or retain the actual text of your prompt.
Think of it as a safety check, not a recording.
- For more information on how Prompt AI processes personal data, please review Section 5.6 of the Team Member Privacy Notice for more details on what information is processed.
26. What about attachments I upload with my prompt?
Attachments are not logged or stored by Prompt AI.
However, attachments are inspected to ensure they do not contain non‑compliant data. If sensitive data is found:
- In some cases, the sensitive data may be redacted, and the prompt can continue.
Example: Certain file types, like .csv files, may allow data to be safely removed and new data inserted.
- In other cases, the upload of the file to the AI tool will be blocked.
Example: If sensitive data is found in an MS Excel (.xlsx) file, the file upload will be blocked because it cannot be safely redacted in an .xlsx file.
This helps prevent accidental sharing of confidential business data or personal data with AI tools.
27. What happens if I accidentally include sensitive information in the text of the prompt?
Prompt AI will do one of two things:
- Redact the sensitive data (remove or mask it) and allow the prompt to continue, or
- Block the prompt entirely if:
- The sensitive data cannot be safely redacted
- The content of the prompt does not align with BW cultural values, for example, prompts containing pornographic, violent, or self-harm content
28. What is logged if my prompt or attachment is redacted or blocked?
If a prompt or attachment is redacted or blocked by Prompt AI, a log entry is created. Information logged includes user name, Prompt AI policy violated, information redacted (as applicable), latency (time to analyze the prompt), IP address from which the prompt originated, country from which the prompt originated (based on the IP address), browser used and version, PC operating system, URL for the AI prompt submission, AI conversation, response and activity ID.
29. Is this used to monitor my performance or behavior?
No, the Prompt AI logs will not be used to monitor performance or behavior.
Prompt AI is focused on information security, it is not aimed at evaluating how AI is used for work tasks.
30. Can Prompt AI see my personal conversations?
- Americas/Middle East/Africa/Asia: Prompt AI inspects all prompts entered into AI tools on BW-managed PCs. Any personal data included in a prompt or submitted as an attachment will be analyzed by Prompt AI. Team members are responsible for carefully evaluating the data they input into AI tools on BW PCs and must ensure that all data is handled lawfully, securely, and confidentially.
- Europe: European team members benefit from stronger workplace protections due to local privacy and labor laws. To comply with such laws, BW is required to find, and committed to honoring, the right balance between its legitimate interest in protecting the business and the rights and freedoms of team members, including their right to private life and correspondence. As part of this effort to balance data loss prevention with respect for team members’ rights and freedoms, BW has decided to prohibit personal use by European team members of AI tools on BW PCs.
31. Will this slow down my work or block AI tools?
In most cases, you will not notice Prompt AI at all.
You may see an impact if:
- Sensitive information is detected in a prompt or attachment.
- A prompt or attachment needs to be redacted or blocked to protect data.
32. What types of data are considered non‑compliant?
Examples include (but are not limited to):
- Confidential business information.
- Personal data such as government ID numbers, credit card numbers, other financial information, or employee details.
- Regulated or protected information.
Policies are designed to focus on clear risk, not everyday work.
- For more information on how Prompt AI processes personal data, please review Section 5.6 of the Team Member Privacy Notice for more details on what information is processed.
33. Who can see Prompt AI logs?
Access to logs is restricted to a limited number of system administrators within the Corporate IT department. IT administrator roles are only granted after a rigorous, documented approval process involving multiple levels of authorization. Logs are used for security investigations, such as activity by a compromised account, or to track the spread of malicious software. Logs are also used to improve the performance of the IT network and applications.
34. What should I do if a prompt or attachment is blocked, but I believe it should be allowed?
If this happens, contact your Service Desk. They can review the situation and help determine next steps.
35. Who can I contact if I have questions?
If you have questions or concerns about Prompt AI or AI usage at BW Company, contact your Service Desk.