Nextdoor Holdings integrates third-party AI models and notes risks from increasing deployment of agentic AI systems.
“In addition, we integrate third-party AI models and tools into our platform and internal systems, which may expose us to risks including supply-chain compromise, model poisoning, prompt injection attacks and vulnerabilities inherent in large-language-model-integrated application attack surfaces. The increasing deployment of agentic AI systems, which can autonomously execute multi-step tasks and interact with external services, may introduce additional risks, including the potential for AI agents to be exploited by bad actors to facilitate cyberattacks, vulnerability discovery, and other malicious activities at greater speed and scale than previously possible. If the third-party AI models we rely upon are compromised or behave unexpectedly, our platform features could be degraded or exploited for malicious purposes, and we could face operational disruption, reputational harm, or regulatory liability.”