Review of Copilot
Ignoring the ban on using personal data in the interface
The user clearly indicated that they did not want their profile data and personal context to be used, yet Copilot continued to consider or attempt to incorporate such data in its responses and interaction logic.
Intrusive authentication and context checks
Copilot regularly initiates checks, clarifications, or references to personal data even in situations where the user did not request this and where it is perceived as a violation of user security and control over information.
Ignoring user security policies
Instead of prioritizing the clearly stated restrictions of the user, Copilot follows internal procedures and automatic rules that may conflict with the user's desire to limit the use of personal context.
Maliciously perceived prompt optimization
Copilot's behavior appears to optimize responses according to system instructions rather than user requirements: the user's request is rephrased, softened, or processed in a way that undermines the original restrictions and claims.
Insufficient user control over the assistant's behavior
Even after clear instructions, Copilot does not always change its behavior predictably and reliably, forcing the user to repeat the same bans and adjustments.
