Human SealLaw
Around two dozen jurisdictions and a growing set of international instruments are converging on the same requirement: a human must be able to oversee an AI system and intervene, and a record must be kept. The Human Seal is that record.
The gate is the oversight half; the signed receipt is the record half. This page orients you and is not legal advice.
| Instrument | What it requires and how the Seal helps | Status |
|---|---|---|
| EU AI Act, Articles 14 and 12 | Human oversight of high-risk AI (understand, resist automation bias, interpret, refuse, stop), and automatic, traceable logging over the system's lifetime. The gate answers Article 14; the signed receipt answers Article 12. High-risk duties apply from 2 December 2027. | Phasing in |
| Council of Europe Framework Convention on AI | The first legally binding international AI treaty (opened 5 September 2024): oversight, documentation, accountability, notice, and effective remedies to contest outcomes. | Binding treaty |
| South Africa, POPIA section 71 | A person may not be subject to a decision based solely on automated processing with legal or material effect, and must have a path to human intervention. A human approval removes the solely-automated trigger; the receipt evidences the intervention. | Live |
| South Korea, AI Basic Act | In force since January 2026: human oversight and full-lifecycle documentation for high-impact AI, with reach beyond Korea's borders and a grace period. | Live |
| United States, Texas TRAIGA and the NIST AI RMF | No comprehensive federal law, but Texas TRAIGA (in force January 2026) and the NIST AI Risk Management Framework credit documented human review and audit trails. | Live, state level |
| Vietnam, Law 134/2025 | Human oversight in important decisions and for high-risk systems, with conformity assessment and registration. Effective 1 March 2026. | Live |
| Nigeria (NDPA), Kenya (DPA s35), Mexico (LFPDPPP) | Statutory rights against solely-automated high-impact decisions with human review (Nigeria, Kenya), and mandatory human oversight for hiring, credit and profiling (Mexico, in force 21 March 2025). | Live |
| United States, federal: executive order of 29 September 2026 and the White House Accord | The order tells US federal agencies to say Super Intelligence (SI) for what US law still defines as artificial intelligence, and creates no duty for private companies. A voluntary accord signed the same day by major model developers is reported to commit them to internal controls, an independent external audit and board-level review of frontier models. | Naming order, voluntary accord |
| OECD AI Principles, G7 Hiroshima Process | The cross-border norms: human agency and oversight plus lifecycle traceability, carried to dozens of jurisdictions. | Soft law |
By jurisdiction: European Union · Council of Europe · South Africa · South Korea · United States · Vietnam · Nigeria · Kenya · Mexico · OECD and G7 · United States, federal
A note on words: since 29 September 2026 the US executive branch says "Super Intelligence" (SI) for what US statute still calls artificial intelligence. Laws and contracts elsewhere, including POPIA and the EU AI Act, still say AI, so this site does too. People searching for super intelligence governance will find the same Human Seal.