Salt Report Finds Gaps in API Security as AI Agent Use Grows
Summary
Salt’s 1H 2026 survey of over 300 security leaders finds gaps in API visibility and agent monitoring, while its customer-data analysis says 99% of attack attempts came from authenticated sources and 65% targeted API misconfiguration.
Key points
- In Salt’s survey, 49% of respondents said they could not monitor machine-to-machine traffic, 48% struggled to distinguish legitimate AI agents from malicious bots, and 47% had delayed deployments over API security concerns.
- The survey found 32% of organizations experienced an API security incident in the past year; sensitive data exposure, vulnerabilities, and authentication problems were commonly reported.
- Nearly 90% of respondents said their organizations already use or plan to use AI-assisted development tools; 60% cited limited control over the security of AI models used in development.
- In Salt Labs’ customer-data analysis, 99% of attack attempts originated from authenticated sources and 99% targeted external-facing APIs.
- Salt Labs attributed 65% of attacks in its analysis to OWASP API8, Security Misconfiguration, and 7% to API1, Broken Object Level Authorization.
- The report recommends mapping relationships among LLMs, MCP servers, and APIs, alongside real-time monitoring and context-aware safeguards for AI-assisted development.
Article Details
- Publisher
- Salt Security
- Report Period
- 1H 2026
- Scope
- Survey of security leaders on API adoption, API security, and AI-agent risks; analysis of Salt Labs customer data.
- Sample Size
- Over 300 security leaders for the survey; customer-data analysis sample size not disclosed.
- Key Statistics
- 32% of organizations experienced an API security incident in the past year.
- 47% reported API growth of 51–100% in the past year.
- 49% reported being unable to monitor machine-to-machine traffic effectively.
- 99% of attack attempts in Salt Labs customer-data analysis originated from authenticated sources; 99% targeted external-facing APIs.
- In the Salt Labs analysis, API8 (Security Misconfiguration) accounted for 65% of attacks.
- Recommendations
- Map relationships between LLMs, MCP servers, and underlying APIs in an Agentic Security Graph.
- Implement real-time monitoring and behavioral detection for machine-to-machine interactions.
- Apply context-aware security guardrails in IDEs and CI/CD pipelines.
- Protect the Agentic Action Layer used by autonomous systems interacting with enterprise infrastructure.
- Align security investments with resilience, compliance, and innovation-speed outcomes.