FERPA School Official Exception for AI Tools

What the School Official Exception Actually Requires

FERPA generally prohibits disclosure of personally identifiable information from education records without consent, but 34 CFR 99.31(a)(1)(i)(B) creates a narrow exception for outside parties treated as school officials. To qualify, a vendor or tool must perform an institutional service or function the school would otherwise use its own employees to perform, remain under the direct control of the institution regarding use and maintenance of the records, and be subject to FERPA's use and redisclosure limitations. Institutions are also required to state, in their annual FERPA notification, the criteria they use to determine who qualifies as a school official, including contractors and other outside parties. None of these requirements reference technology type. The test is functional: what the tool does with the data and who controls that use, not how the tool is built.

Why AI Agents Complicate the Direct Control Analysis

Direct control has historically been satisfied through contract language: purpose limitation clauses, security requirements, and data destruction or return obligations. That approach assumed a relatively static integration, where a vendor's system processed data in predictable, auditable ways. Agentic AI systems introduce a different pattern. An AI agent can autonomously decide which tools to call, which records to retrieve, and how to combine or forward that data across multiple steps or sub-systems. This tool-call decision layer did not exist in the vendor relationships that current FERPA guidance was written around. The legal standard has not changed, but the evidence required to satisfy it has. A contract stating that a vendor will limit use of student data is no longer sufficient on its own if the underlying system can dynamically expand its own data access or route information to services the contract did not anticipate.

No AI-Specific FERPA Guidance Exists Yet

There is no confirmed Department of Education, FPCO, or Student Privacy Policy Office guidance, rulemaking, or enforcement action that specifically addresses AI agents or automated tool-calling under FERPA. Existing SPPO guidance on online educational service providers predates the current generation of agentic and multi-step AI tools. This leaves an interpretive gap: institutions and vendors must apply a regulation and guidance written for simpler vendor relationships to systems capable of autonomous, multi-hop data access. In the absence of AI-specific rulemaking, the burden falls on institutions to demonstrate, through contracts and technical evidence, that direct control and use limitation are actually being enforced, not just documented.

Governance Implications for Institutions and Vendors

Legitimate educational interest must be tied to a defined institutional function. A broad or open-ended AI use case weakens the school official argument because it becomes harder to show the tool is performing a specific, bounded service the school would otherwise staff internally. Use limitation requires demonstrable enforcement, not just policy language, once an agent can dynamically determine what data to access or how to use it. Institutions should assign clear accountability for verifying, on a recurring basis, that vendor practices continue to satisfy direct control and use limitation, particularly as vendors update model capabilities or add new tool integrations. Multi-vendor AI stacks add further complexity: if an underlying model provider is separate from the primary EdTech vendor, each party's role in processing education records may need independent evaluation against the same criteria.