Guest: Tim Erlin
Company: Wallarm
Show: The Agentic Enterprise
Topic: AI agents

APIs have quietly become the single most exploited attack surface in 2026. As organizations race to adopt AI agents, automation, and intelligent systems, a critical truth is emerging: every AI interaction, every automated decision, every agent-driven action flows through APIs. For CISOs treating API security as just another compliance checkbox, the latest data suggests it’s time for a major strategic shift.

Tim Erlin, Security Strategist at Wallarm, recently shared insights from their latest API Threat Stats Report—the 15th such report since 2022—revealing both persistent threats and alarming new trends. The data tells a clear story: AI security and API security are inseparable, and the explosion of AI is creating unprecedented attack surfaces.

The Numbers Behind the Threat

Wallarm’s analysis of real attack data shows remarkable consistency in attacker behavior. Injection attacks, particularly SQL injection, continue to dominate the threat landscape, holding the number two position in their API Threat Stats Top 10. Authentication and authorization issues occupy positions three and four. “Attackers continue to focus on the easiest-to-exploit types of attacks at scale,” Erlin explains. “Those attacks may or may not be successful, but they continue to drive their campaigns with these easy-to-scale, not-too-complex attacks.”

But the more alarming trend is the intersection of AI and API vulnerabilities. According to the report, 36% of AI vulnerabilities published in 2025 are also API vulnerabilities. Even more striking, AI vulnerabilities themselves grew by approximately 400% year over year. “We see this incredible growth in terms of AI risk in the industry, and AI risk and API risk are tightly connected because you can’t deploy and run AI without APIs,” Erlin notes.

The Shadow Problem: APIs and AI

The challenge facing organizations mirrors familiar security problems, but at a new scale. Just as zombie APIs and shadow APIs have long plagued security teams, shadow AI is emerging as the next frontier. “Organizations don’t have a handle on what AI is deployed in their environment,” Erlin warns. “They don’t have visibility into what AI is deployed in their environment.”

This lack of visibility is particularly dangerous as AI adoption accelerates faster than security controls can keep pace. While organizations are gradually improving their understanding of API risks and implementing stronger protections, the rapid expansion of AI is outpacing that progress. “While we’re improving around API security, we’re really working hard to catch up on the AI side of things,” Erlin acknowledges.

Why AI Security IS API Security

The connection between AI and APIs isn’t just conceptual—it’s architectural. Every interaction with AI tools, whether chatbots, large language models, or AI agents, is mediated through APIs. When those AI tools take actions, those actions are also executed through APIs. “Not only do the AI tools that we’re deploying depend on a foundation of APIs, they’re also driving an increase in the number of APIs that exist to support them,” Erlin explains.

This creates a paradox for security teams. As organizations become “AI native,” APIs recede into the background, handled automatically by AI layers. For developers focused on AI, the API infrastructure becomes almost invisible. But for security practitioners, this means protecting an expanding attack surface that’s increasingly abstracted away from view. “We have to think about the risk across the entire stack, from the APIs up to the AI layer,” Erlin emphasizes.

From Tech Problem to Business Risk

One of the report’s key messages is that API security has evolved from a technical concern to a business-critical issue. CISOs are increasingly able to draw direct lines between API security and revenue protection, partly because APIs often have clear connections to revenue-generating services.

“When you look at the infrastructure that you’re protecting, you’re protecting it in order to allow the organization to continue delivering its product or its service that it sells in order to generate revenue,” Erlin says. “APIs often have a very direct tie to revenue, so it’s a clear connection—I need to protect this API because if it’s compromised, we lose revenue.”

This business-focused framing helps CISOs communicate with boards in language they understand: risk tolerance and financial impact. The challenge isn’t achieving perfect security—it’s supporting the organization’s mission while managing acceptable levels of risk.

The New Risk Boundary: Behavior

Beyond high-volume, low-sophistication attacks, the report identifies a more concerning trend: the rise of behavior-based and business logic attacks. Rather than exploiting code vulnerabilities, attackers are increasingly leveraging the way applications are designed to work. “Where we see the most interesting developments in terms of API attacks is around more business logic, behavior based attacks,” Erlin notes. “Attackers are leveraging the way an application works, as opposed to vulnerabilities in the code.”

This shift represents a fundamental change in the risk boundary. Traditional security controls focused on preventing known exploits may miss attacks that use legitimate functionality in unintended ways.

Moving Forward

For organizations navigating the AI transformation, Erlin offers clear guidance: understand that AI security is API security, recognize that this is a business risk requiring board-level attention, and prepare for behavior-based attacks that go beyond traditional vulnerability management.

Wallarm’s approach to these challenges includes API discovery across environments, inline detection and blocking of both traditional and business logic attacks, security testing before production deployment, and governance capabilities for policy enforcement and API specification management.

As Sam Altman noted, every company is becoming an API company. The question isn’t whether organizations will adopt APIs and AI—it’s whether they’ll secure them effectively enough to protect their business.