An organization has created a specialized AI service that processes requests from external sources and provides predictive outcomes. The team observes unusual query patterns suggesting a systematic effort to replicate the underlying logic. Which measure best addresses this concern?
Increase the size of the processing hardware so the entire environment changes daily
Deploy a separate online environment where every user must have a unique client certificate to connect
Disable all logging services to prevent visibility into operational data
Implement usage throttling paired with techniques that mask sensitive details of the response
Combining usage throttling with techniques that alter output data makes it harder to collect enough information to replicate the logic. Throttling limits large-scale interactions, and output obfuscation complicates any attempts to reverse-engineer the system’s underlying parameters. Other options address security or infrastructure concerns but do not specifically mitigate repeated data collection for the purpose of reconstructing the algorithm.
Ask Bash
Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.
What is usage throttling, and how does it help protect an AI service?
Open an interactive chat with Bash
What is output obfuscation, and why is it important for protecting AI models?
Open an interactive chat with Bash
Why are options like disabling logs or changing hardware ineffective against reverse-engineering attempts?