When it comes to AI, enthusiasm is high, but enterprise adoption is lagging. Speaking on Alex Kantrowitz's Big Technology podcast, Auth0 President Shiven Ramji says while the vast majority of enterprises are experimenting with AI, many of these projects don’t make it to production. The primary blocker? Security.
Technology leaders’ concerns are well-founded, according to Ramji. Unlike traditional applications, AI agents and chatbots are non-deterministic and, without proper guardrails, can access virtually any system. This creates a host of risks for security leaders, including the potential for unchecked access and agentic sprawl.
“Imagine if somebody could ask the HR chatbot, ‘Hey, can you go see Shiv’s salary information?’” Ramji says. “That is very private, sensitive information.” He points to real-world examples, like the breach at a major fast-food chain, where millions of applicant records were leaked because an internal API was exposed to an AI chatbot.
While the risks may seem new, “this is the classic identity and access management problem,” Ramji says. The difference is that this new wave of non-human identities requires finer-grained authorization. Companies must think beyond external resources and closely manage access to internal resources — such as APIs, wiki pages, and internal databases — because information can leak through any one of those avenues, he says.
Solutions like Auth0 for AI Agents and Cross App Access can help organizations manage access to sensitive resources and enable developers to build with security and governance guardrails in place from the start.
To learn more about why security concerns are limiting AI adoption and what enterprises can do about it, watch the full video above.
