By IT Brew Staff
less than 3 min read
Definition:
Although AI can prove a powerful tool for a variety of functions and industries, it also has the potential to introduce serious legal and ethical issues. For example, without the proper guardrails in place, an AI chatbot could accidentally reveal sensitive customer data or vulnerabilities in the company’s code. Thoughtful AI governance can help mitigate these issues by introducing oversight into every part of the AI development life cycle, from training AI models to the output of software tools built with AI.
At its best, AI governance includes mechanisms that delegate responsibility to the proper stakeholders throughout the organization, while insisting on transparency of AI systems and data. If things go wrong, stakeholders also need a risk management plan for dealing with the organizational, ethical, and reputational repercussions.
With enough time and resources, any organization can establish its own AI governance panel or working group. These entities often establish a code of AI ethics, talk to AI stakeholders, and communicate the governance principles to the broader organization. While AI governance may look different between organizations, it often aligns to a few key pillars, which could include:
- Transparency: Does the organization have visibility into the AI tool’s data and why it produces certain outcomes?
- Accountability: Are there stakeholders accountable for the AI tool’s behavior? Does the AI comply with laws and regulations?
- Privacy: Does the AI tool keep sensitive data private?
- Security: Is there security in place to ensure the AI tool isn’t a potential attack vector?
- Business objectives: Does the AI tool fulfill the organization’s business objectives, and does it do so in a way that’s scalable and sustainable?
Organizations may also turn to pre-established data and AI frameworks such as the General Data Protection Regulation (GDPR) and the NIST AI Risk Management Framework to help guide their thinking and procedures.
