AI in IAM

02 May, 2023, by Rachel Swann

When I sat down to write this article, I was immediately stumped on how to open the conversation about AI without making people assume the worst. In days gone by, the average person heard AI and thought of Skynet, HAL 9000, or Agent Smith. Evil but fictional; not something to lose sleep over. Now the average person thinks of losing jobs, stolen art, or fake defamatory videos. A very different evil is at the forefront of the AI debate, one that only serves to terrify people not in the know. So, I had a problem: how do you talk about the benefits of AI for identity and access management when so many people have already made up their minds about the concept of AI?

My solution was simple. Instead of me fighting on the frontlines, I can just have AI defend itself. Here's what ChatGPT had to say for itself:

"Artificial Intelligence (AI) has the potential to revolutionize Identity & Access Management (IAM) by providing more efficient and effective ways to manage access to digital resources. Here are some compelling reasons why AI can be a good thing for IAM:

• Predictive modeling: You could use AI to build a predictive model that identifies potential security risks based on patterns in the data model. For example, the AI might detect a pattern of accounts being granted permissions that are not typically associated with a particular role, or an identity being associated with multiple accounts that are not typically associated with that identity.
• Anomaly detection: Another approach would be to use AI to identify anomalies in the data model. For example, the AI might detect an identity that has been granted permissions that are not typically associated with that identity or an account that has been granted permissions that are not typically associated with that account.
• Recommendation engine: You could also use AI to build a recommendation engine that suggests changes to the data model based on patterns in the data. For example, the AI might recommend that certain permissions be removed from a role or that an identity be associated with a different account.
• Natural language processing: Finally, you could use AI-powered natural language processing (NLP) to analyze user requests for changes to the data model. The NLP could analyze the request to identify the relevant identities, roles, permissions, and accounts, and then make recommendations for changes based on the data.

In conclusion, while AI can offer many benefits to IAM, it's important to recognize that it is not a panacea and must be used appropriately and with caution. Organizations should carefully consider the potential drawbacks and limitations of AI and ensure that it is implemented in a way that is transparent, fair, and aligned with their broader goals and values."

A strong rebuttal from ChatGPT, outlining just what makes AI useful for identity management. I found myself agreeing with ChatGPT on a lot of these points. There's just some things that are easier for a robot to do, like repetitive tasks. AI doesn't get bored and miss subtle mistakes, like a person does. But AI is not a perfect solution, and there are jobs better left to those not made of code, as ChatGPT continues:

"While using artificial intelligence to analyze data models can be a useful tool for gaining insights, there are some potential drawbacks to consider. For example:

• Data Bias: AI systems are only as good as the data they are trained on. If the data is biased or incomplete, the AI model may generate inaccurate or misleading insights.
• Lack of Transparency: AI models can be complex and difficult to understand. It can be challenging to identify why the model arrived at a particular conclusion, which can be problematic for decision-making.
• Limited Context: AI models may lack the contextual understanding necessary to make accurate predictions or identify patterns. For example, an AI system may not be able to account for the impact of external events or changes in the business environment.
• High Cost: Implementing AI systems can be expensive, requiring significant investment in hardware, software, and personnel. This cost may not be justifiable for smaller organizations or projects.
• Overreliance on AI: Relying too heavily on AI models to make decisions can lead to complacency or a lack of critical thinking. It is important to balance the insights generated by AI with human intuition and expertise.

In conclusion, while AI can be a powerful tool for analyzing data models, it is important to consider the potential drawbacks and limitations of using these systems. Organizations should weigh the costs and benefits of implementing AI and ensure that they have the necessary data and expertise to use these systems effectively."

Ultimately when it comes to AI there’s no such thing as good or bad. AI is only as good as the data you train it on, and that data will have messy human fingerprints all over it.

ChatGPT comes up with a good point in the end: It is important to balance the insights generated by AI with human intuition. Like this article, the future of AI in IAM is dependent on how we use it, how we collaborate to make something great.