Wallarm warns APIs are fueling AI-era breaches at machine speed – News

A new report out today from application security company Wallarm Inc. warns that application programming interfaces have become the single most exploited attack surface across vulnerabilities, active exploits and real-world breaches.

The Wallarm 2026 API Threat Stats Report — the New API Risk Multiplier is based on analysis of 67,058 published vulnerabilities in 2025, the U.S. Cybersecurity and Infrastructure Security Agency’s Known Exploited Vulnerabilities catalog and 60 publicly disclosed API-related breaches. It found that APIs are no longer just part of the application security conversation but are now the dominant surface attackers target.

Wallarm found that 11,053 vulnerabilities published in 2025 — 17% of the total — were API-related. Of the 245 vulnerabilities added to CISA’s KEV list during the year, 43% were API-related. In the latter, APIs were the single most common exploited surface in the dataset.

Not surprisingly, the convergence of artificial intelligence and APIs was found to be accelerating the problem.

The company identified 2,185 AI-related vulnerabilities in 2025, with 786 overlapping API vulnerabilities, meaning 36% of AI flaws involved APIs. The same 36% overlap appeared in exploited AI vulnerabilities, reinforcing the idea that “AI security is API security” in practice, not just theory.

The report details how the characteristics of API flaws make them particularly dangerous, with 97% of API vulnerabilities able to be exploited with a single request, 98% rated easy or trivial to exploit, and 99% remotely exploitable. Moreover, in 59% of cases, no authentication is required.

The result, according to the report, is an attack surface optimized for speed, automation and scale rather than sophisticated, multistep intrusion.

Attack telemetry was also found to show a shift in attacker behavior. In Wallarm’s ThreatStats Top 10, “Cross-Site Issues” rose to the top category by observed attack volume in 2025, overtaking injection flaws, which nevertheless remained a persistent high-impact threat. Broken access control and insecure resource consumption continued to enable large-scale abuse.

Other findings in the report include that the Model Context Protocol has quickly emerged as a growing risk, with Wallarm identifying 315 MCP-related vulnerabilities in 2025 — 14% of all AI vulnerabilities. MCP-related flaws were also tied to a Top 10 API breach involving thousands of exposed MCP servers, highlighting how APIs that act on behalf of autonomous agents can amplify the blast radius of a single control failure.

“API security is at the heart of any AI transformation,” said Ivan Novikov, founder and chief executive officer of Wallarm. “Every AI application or agent interaction is mediated through an API. API security is integral to successful AI adoption and AI by its very nature has made the consequences of getting it wrong much larger and much more impactful.”

The report concludes by noting that for security leaders, the takeaway is direct: Improving API security is not about chasing new attack classes. It’s about systematically addressing identity, exposure and abuse before automation and scale turn familiar weaknesses into material business risk.

Image: News/Ideogram