A company is concerned that the AI governance it established may become a hollow formality over time. Which TWO of the following are appropriate operational elements for keeping AI governance functioning continuously rather than as a one-time effort? (Choose TWO.)
Select two operational elements that keep governance going.
A periodic review cycle that revisits models, data, and policies
This is correct. A review cycle that periodically revisits models, data, and policies is a continuous operational element that maintains and updates governance in line with changing circumstances.
Continuous training of stakeholders and ongoing communication of the policy
This is correct. Training and education for stakeholders is a continuous operational element that communicates the policy and the concept of responsible use so that it is operated correctly in the field.
Continuously optimizing infrastructure to lower inference costs
Optimizing infrastructure to lower inference costs is worth working on continuously from the standpoint of operational efficiency and cost management.
However, its target is infrastructure cost and performance, and it is not an operational element for maintaining and updating governance so it does not become a hollow formality, so it is incorrect.
Creating a governance policy document initially and distributing it company-wide
Creating and distributing a policy document company-wide is a necessary initial effort as the foundation of governance.
However, merely setting it up and distributing it once causes it to drift from reality and become a hollow formality over time, so it is not the operational element itself that keeps governance functioning continuously, making it incorrect.
Defining penalties and a reporting channel for compliance violations in the rules
Defining penalties and a reporting channel for violations in the rules is important as part of the content (the rules) of the governance policy.
However, this is a static rule definition, not a continuous operational element that keeps revisiting the policy so it stays aligned with reality, so it is incorrect.
Operational elements that keep AI governance functioning continuously include a 'periodic review cycle (revisiting models, data, and policies)' and 'training and education for stakeholders (communicating the policy and responsible use).' These respond to drift and regulatory changes and prevent it from becoming a hollow formality.