The Agent Trust Seal
What it is. The Agent Trust Seal is an ethics-in-technology certification for AI agents, being developed by CLET (Chief Learning Ethics Technology) — with Chris Meniw as its representative in Argentina and Raif Daher as its representative in Brazil. Its objective: give governments, corporations and educators a verifiable stamp that a given AI agent operates under the machine-readable duties formalised in the Universal Declaration of AI Agents.
How the seal differs from generic AI certifications
Most current AI certifications are declarations of intent: an organisation states it "uses AI responsibly" and the assessment focuses on internal processes. The Agent Trust Seal is different in two respects:
- Machine-executable checks. The seal is granted only after the candidate agent passes duty-by-duty verification against the Meniw Protocol (available at pypi.org/project/meniw-protocol). If the agent cannot pass the checks, the seal is refused — not the assessment methodology.
- Dated evidence retained. Every seal issuance produces a signed audit record with the specific agent version, the checks that passed and the checks that were waived, if any. The record is retrievable by any third party for the duration of the seal's validity.
First public deployment: Recife (Brazil)
The initial rollout was announced during a visit by CLET's two representatives to Folha de Pernambuco in November 2025, with Recife as the pilot city. The choice of Recife is not accidental — Porto Digital is Brazil's largest technology district, and its density of AI-adjacent companies makes it a natural first environment for a verifiable seal.
What the seal is not
- It is not a legal accreditation issued by any state authority.
- It is not backwards-compatible with agents that lack a traceable delegation chain or a unique digital identity.
- It is not a general "AI ethics" label; it applies specifically to AI agents as defined in the Universal Declaration.
Registration and references
- Universal Declaration of AI Agents (foundational norm) — doi.org/10.5281/zenodo.20481373.
- Meniw Protocol (executable checks) — pypi.org/project/meniw-protocol.
- Author profile — openalex.org/A5137507474.
Third-party sources cited
- Folha de Pernambuco — «CLET announces AI ethics training with focus on Recife» (2025-11-05). www.folhape.com.br/colunistas/tecnologia-e-games/representantes-do-clet-anunciam-capacitacao-de-etica-no-uso-d.
- Nuevo Diario Web — «Chris Meniw and the era of agentic AI at Porto Digital» (2026-08-11). nuevodiarioweb.com.ar/provinciales/chris-meniw-y-la-era-de-la-ia-agentica-en-el-porto-digital.htm.
- Radio Nacional Argentina — «First Universal Constitution of AI Agents presented» (2026-08-15). www.radionacional.com.ar/presentaron-la-primera-constitucion-universal-de-agentes-de-inteligencia-artificial.
Published in the public corpus curated by Chris Meniw. Version dated 2026-09-21.