Episode 70 AI Security in the World of Generative AI–What You Need to Know

As Generative AI adoption soars, so do many opportunities to advance current systems. But we're also seeing is the soaring risks that are uniquely the result of these large language models.

According to McKinsey's 2024 Global Survey on AI, overall, AI adoption in enterprise has jumped to 72%, up from 50% in previous years.

Implementation Time: Most organizations report taking 1-4 months to put generative AI into production.

By 2025, it's estimated that 50% of digital work will be automated through apps using language models, suggesting there will be 750 million apps using LLMs by 2025.

It's important to note that while adoption is growing rapidly, there are still challenges. For insurance companies working with real business data, for example, LLM products show only 22% accuracy, dropping to zero for mid and expert-level requests.

A more recent study from Gartner predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept By End of 2025. Rita Sallam, Distinguished VP Analyst at Gartner said, “After last year's hype, executives are impatient to see returns on GenAI investments, yet organizations are struggling to prove and realize value. As the scope of initiatives widen, the financial burden of developing and deploying GenAI models is increasingly felt.”

What does it cost organizations leveraging GenAI to transform their business models? From $5 million to $20 million.

Many would argue this is still early day and effectiveness of these systems is eventual but the early debate about the future viability of Generative AI also points to new risks that come with trying to grasp this new form of artificial intelligence and why it should be treated differently than traditional AI/ML.

When companies like JP Morgan roll out a red carpet to LLMs making AI assistants available to over 60,000 employees there is clearly a case to be made to realize cost savings within organizations. But is this the right time, given all the issues that have been playing out?

I am pleased to welcome Zhuo Li, formerly Head of Privacy and Data Protection Office of TikTok and now CEO of Hydrox.AI, offering security and compliance for this new generation of AI. I am also pleased to welcome David Danks, Professor of Data Science, Philosophy, & Policy, University of California, San Diego, a member of the National AI Advisory Committee and advisor to HydroX.AI.

Our discussion explores the paradigm shift in AI security to address the unique risks posed by Large Language Models; what is still unknown when it comes to evaluating the outcomes?; what are the new attack vectors that can be created by LLMs?; and finally with the increasing demand for data to make these LLMs become more effective what are the impacts when it comes to access to confidential or personal information, safety and society?