Washington, D.C. — 25th April 2025 — FireTail, the leading AI & API security platform, has released its annual report, The State of AI & API Security 2025, revealing a critical blind spot in the way organizations are securing their AI investments. Despite record-breaking AI adoption, the report warns that most enterprises are overlooking the most exposed part of the AI stack: the API layer.

“APIs are the foundation of AI applications, and attackers know it,” said Jeremy Snyder, Co-founder and CEO at FireTail. “If you don’t secure your APIs, you’re not securing your AI. It’s that simple.”

The report is based on research and analysis from FireTail’s API Breach Tracker, its AI Incident Tracker, telemetry from production environments, and detailed reviews of major AI-related security incidents. It provides concrete evidence that APIs are not just enabling AI, but exposing it.

Key Insights from the Report

In 2024, FireTail tracked 26 major API security incidents, an increase from 22 in the previous year. It showed that attackers continue to exploit long-standing vulnerabilities such as authorization flaws, weak authentication, and insufficient input validation. Since 2017, more than 1.6 billion records have been exposed via API-related incidents.

The report also examines high-profile cases including the Irish Government’s vaccination portal, OpenAI’s web crawler, and Meta’s LLaMA framework.

“We’re already seeing the first wave of large-scale AI breaches,” Snyder said. “And it’s clear that organizations are rushing to adopt AI without proper security oversight.”

A Growing Attack Surface and A Lack of Visibility

Research shows that 97% of organizations believe AI introduces unique security challenges, yet nearly 60% say they lack visibility into the APIs powering their AI systems. This gap is creating opportunities for attackers to exploit shadow APIs, bypass controls, and launch techniques like prompt injection and model poisoning.

The report also highlights the regulatory shift taking place. The FCC’s $16 million enforcement action against TracFone over API vulnerabilities shows that regulators now view API security failures as compliance violations rather in addition to being  technical and security oversights.

Frameworks like the CIS API Security Guide, OWASP LLM Top 10, and ISO 42001 are beginning to provide much-needed structure. However, FireTail argues that these must be paired with proactive discovery, posture management, and runtime protection to be truly effective.

A Call for API-First Security

The State of AI & API Security 2025 calls for an API-centric approach to securing AI. Key recommendations include comprehensive API discovery, strong authentication and authorization, secure-by-design development, and continuous monitoring.

“AI doesn’t exist in a vacuum. It’s connected to everything, and that connection point is the API,” said Snyder. “If organizations want to innovate safely with AI, they need to start by securing the APIs that power it.”

Access the Report

The State of AI & API Security 2025 is available for download at:
firetail.ai/reports/the-state-of-ai-and-api-security-2025

‍

‍