We use analytics cookies to understand how readers use the site. They only run if you accept. Privacy policy.
When a safety-focused AI lab deploys tools inside the NSA under legal authorities that predate machine learning, accountability becomes almost impossible to locate.
| Framework | Enacted | Governs | AI Provisions | Binding on AI |
|---|---|---|---|---|
| Executive Order 12333 | 1981 | NSA signals intelligence collection | None | No |
| Foreign Intelligence Surveillance Act (FISA) | 1978 | Domestic intelligence collection; FISA Court oversight | None | No |
| USA FREEDOM Act | 2015 | Limits on bulk phone metadata collection post-Snowden | None | No |
| EO 14110 — Safe, Secure, Trustworthy AI | 2023 | Federal agency AI use; safety standards; reporting requirements | Yes | Partial |
| No comprehensive federal AI statute | — | N/A — does not exist | None | No |
Note: No row in this table represents a framework that is both AI-specific and fully binding on classified intelligence community deployments.
Hover any event to read its description. Circle size indicates significance. Legal frameworks (orange) anchor the left; the Mythos deployment (far right) lands in a space none of them were designed to reach.
When a safety-focused AI lab deploys tools inside the NSA under legal authorities that predate machine learning, accountability becomes almost impossible to locate.
The Senate Select Committee on Intelligence, the House Permanent Select Committee on Intelligence, and the Senate Judiciary Committee each hold partial oversight authority — but none has enacted AI-specific binding rules for the intelligence community. The Mythos deployment is a test case for whether accountability can exist in classified AI at all.