Responsible Innovation in Generative AI ― Navigating Ethics, Compliance, and Governance

As generative AI transforms industries, IT and security leaders must take the lead to ensure responsible development and deployment. However, with each new use case, ethical dilemmas and compliance concerns arise. Can sensitive data be protected? Can AI be used in a way that respects copyrights? 

Evanta CIOs in the Southern California Community got together recently to discuss how to operationalize and scale AI, manage and mitigate the risks, and navigate new AI regulations and the associated data security challenges.

IBM and Adobe hosted the discussion on “Responsible Innovation in Generative AI ― Navigating Ethics, Compliance, and Governance,” which was moderated by Tim Richer, Director, Product Marketing, Data and AI at IBM. IBM and its alliance partners have worked closely with customers to tackle governance challenges and outlined pitfalls to avoid and opportunistic strategies to consider to leverage AI in a responsible way.

Evanta community members Deanna Steele, CIO at Allied Universal; Gopal Govindarajulu, Vice President Information Technology at Enlyte; Ashish Srivastava, Vice President - Digital & IT (CDIO) at Daikin Applied; and Louis Zhan, Senior Vice President, Enterprise Solutions and Services at Beachbody, served as discussion leaders.
 

Five Truths of Generative AI

Tim Richer of IBM set the stage for the discussion, sharing the five common challenges they have observed after a year and a half of working with clients on implementing Generative AI.

1. Model discovery. Organizations have to determine which model is right for each use case. Some are leveraging general purpose models or consumer models and combining them with proprietary models that organizations are building themselves.
2. Multi or hybrid cloud. Another factor is where to run an organization’s AI systems and platforms. Companies that have private/public clouds, or a multi-cloud approach, have to figure out how to co-locate their AI platform with their data, which is a challenge at scale.
3. Establishing governance. Companies need not only data governance, but AI governance in order to ensure the models and systems are doing what they want them to do. 
4. Scaling for value. Executives are currently asking if they are selecting the right use cases and if they are investing in the areas where they are going to have maximum impact.
5. Data matters. Implementing GenAI quickly becomes a data conversation. Organizations should leverage their data to refine AI and use their proprietary assets as much as they can.

In addition to these five common challenges, Tim noted the other obstacles to widespread GenAI adoption, including a rapidly evolving regulatory landscape, the complexity of organizational dynamics and multiple stakeholders, and determining how to manage AI governance. 

On the topic of governance, Tim shared two models for CIOs to consider. He explained that Organizational AI Governance is the overarching strategy, use cases, stakeholders involved in the use cases, and the market implication of the use cases. IT leaders should also look at an Automated AI Model Lifecycle Governance, which he described as the “through model.” This model extends from the strategy to integrating it into the system, training the data, and integrating it into a workflow. It can also include measurement and a feedback loop of measuring the model. 
 

Key Takeaways from the Discussion

During breakout groups, CIOs discussed GenAI governance, collaborating on use cases and enforcing policies across their organizations.

• Establish a GenAI governance framework. CIOs believe that establishing a governance framework before deploying generative AI is crucial. This includes defining acceptable use cases and ensuring regulatory compliance. Some executives noted that their organizations are forming AI councils or ethics committees to discuss their organization’s "risk appetite" and determine what use cases should move from brainstorming to actual implementation.
• Enforce governance policies across the organization. CIOs in the discussion also shared that policies on AI need to be enforced across the organization to ensure compliance and ethical use. One executive mentioned that they are implementing AI policies within their cybersecurity tools, which prompt or nudge users if they do something outside the policy.
• Involve your Compliance and Legal teams. CIOs are leaning on their Compliance and Legal teams to help navigate the complex regulatory landscape and ensure that AI initiatives meet legal requirements. One executive said, “It’s challenging to do this, but necessary.” Some IT leaders that are farther along in AI production are more likely to be including Compliance groups to help manage AI risks effectively.
• Collaborate cross-functionally with other business leaders. IT leaders emphasized the need to establish cross-functional discussions when it comes to GenAI implementation. Bringing together business and IT teams to discuss AI use cases helps in understanding the tools needed and educating the organization about the risks and benefits of AI. One CIO mentioned that their company regularly convenes business and IT stakeholders to explore AI use cases, and at the same time, emphasize that “AI is a double edged sword” with both risks and opportunities.
• Prepare to adapt to rapid technological changes. With the fast-evolving AI landscape, CIOs need to consider the longevity of AI tools and the length of contracts with AI vendors. Executives said they are looking at both small and large partners and evaluating contract lengths to ensure they are not locked into outdated technology as the landscape changes.

One CIO summed up their approach to GenAI governance by saying, “Most of us do have a governance process and structure in place, including legal, security and data. We are just at different stages of maturity.” 
 

CIOs can continue the conversation on navigating the implementation of Generative AI at an upcoming Evanta community gathering. If you are already a community member, sign in to MyEvanta to register, or you can apply to join a community of your CIO peers to stay fully up-to-date on critical topics for IT leaders.
 

Special thanks to IBM and Adobe.

by CIOs, for CIOs
 

Join the conversation with peers in your local CIO community.