SAS: Organisations Lack Measures to Tackle AI, Data Bias

April 18, 2024

undefined mins

Research by SAS has found that 80% of leaders are concerned about data privacy and security

Organisations are enthusiastic about Gen AI’s potential for increasing their productivity, but SAS research finds data privacy concerns remain

The meteoric rise of generative artificial intelligence (Gen AI) models like ChatGPT has opened up exciting new possibilities for businesses across industries. These powerful language models can assist with a wide variety of tasks, from content creation to coding to customer service. However, the rapid adoption of Gen AI is also exposing businesses to significant data privacy risks that require urgent attention.

A key concern is the training data used to develop these AI systems. Large language models ingest massive troves of online information, including websites, books, articles, and social media posts – much of which can contain personal data, intellectual property, and other sensitive information. Businesses that leverage these models must grapple with the complex legal and ethical implications of using data that was never intended for this purpose.

Every day, the world produces five exabytes of data. By 2025, this is set to rise to a rate of 463 exabytes per day, driven in part by increased adoption of Gen AI. But as businesses continue to embrace AI at an accelerating pace, organisations today are increasingly concerned about how the technology could affect their valuable data.

Now, research by SAS announced at its SAS Innovate event has found that 80% of leaders are concerned about data privacy and security, with business leaders admitting a lack of governance frameworks.

The report found that US organisations are enthusiastic about Gen AI’s potential for increasing their business and people productivity. But beneath surging enthusiasm, leaders see understanding gaps, a lack of strategic planning and the talent famine as obstacles to realising and measuring the technology’s full value.

Challenges faced by organisations implementing Gen AI

Organisations are facing several key challenges as they work to implement Gen AI. Firstly, they are struggling to increase trust in their data usage and achieve compliance. The SAS report found only one in 10 organisations has a reliable system in place to measure bias and privacy risk in large language models (LLMs), and a staggering 93% of US businesses lack a comprehensive governance framework for Gen AI, putting the majority at risk of noncompliance with emerging regulations.

Secondly, organisations are encountering compatibility issues when trying to integrate Gen AI into their existing systems and processes. The seamless integration of these new technologies with legacy infrastructure remains a significant hurdle.

Another third challenge involves talent and skills. Organisations are finding that in-house GenAI expertise is severely lacking, as HR departments encounter a scarcity of suitable hires. Organisational leaders worry they simply do not have access to the necessary skills to fully capitalise on their Gen AI investments.

Finally, predicting the costs associated with using LLMs has proven to be a major challenge. While model creators provide initial cost estimates, leaders cite prohibitive direct and indirect expenses related to private knowledge preparation, training, and model operations management. The true financial implications of GenAI deployment are complex and often underestimated.

“Organisations are realising that large language models alone don’t solve business challenges,” said Marinela Profi, Strategic AI Advisor at SAS. “Gen AI should be treated as an ideal contributor to hyper automation and the acceleration of existing processes and systems rather than the new shiny toy that will help organisations realise all their business aspirations. Time spent developing a progressive strategy and investing in technology that offers integration, governance and explainability of LLMs are crucial steps all organisations should take before jumping in with both feet and getting ‘locked in.’

“It’s going to come down to identifying real-world use cases that deliver the highest value and solve human needs in a sustainable and scalable manner. Through this study, we’re continuing our commitment to helping organisations stay relevant, invest their money wisely and remain resilient. In an era where AI technology evolves almost daily, competitive advantage is highly dependent on the ability to embrace the resiliency rules.”