Certification marks and trust registries

As more services run through an AI agent, and agencies come to rely on more third-party tools, the marks and claims those tools display multiply faster than anyone can verify them. An agent has nothing it can check programmatically at all. Physical goods have CE marks, electrical safety tags, and food-grade certifications; digital tools have no equivalent that is both widely recognized and independently assured.

Emerging. The public registry and trust-badge response is in operation but narrow: a machine-readable mark an agent can query is legislated under CE marking yet not operationally implemented, FedRAMP-style certification is mature for cloud but only beginning to cover AI, and the DPGA registry is live but voluntary and limited in scope.

EU AI Act Article 48, CE marking for high-risk systems. Providers of high-risk AI systems must undergo conformity assessment and affix a CE marking, the symbol long used for physical product safety, and for systems provided digitally the Act requires a digital CE marking accessible through the software interface or a machine-readable code. Where a Notified Body conducted the assessment, its identification number must follow the mark. Article 79 carries the withdrawal power: market surveillance authorities may require an operator to bring a system into compliance, withdraw it, or recall it.

EU AI Act Article 43, who performs the assessment. For the high-risk categories at points 2 to 8 of Annex III the Act routes a provider to conformity assessment based on internal control, with no notified body involved. A notified body enters at point 1, and there only where harmonised standards were not applied or were applied in part. The mark records the provider's own assessment, across most of the scope, under a procedure the reader of the mark never sees.

The FedRAMP Marketplace (US). The registry lists certified cloud service offerings alongside the agencies that certified them and the assessors the program recognizes. FedRAMP's own name for the model is do once, use many: a provider is certified once, and the resulting security package can be reused by any federal agency. GSA has since prioritized 20x authorizations for AI cloud services, beginning with conversational engines already on the government schedule, so a registry model matured on cloud is reaching AI one service category at a time.

The Digital Public Goods Alliance registry. The registry lists verified digital public goods assessed against a nine-indicator standard covering SDG relevance, approved open licenses, clear ownership, platform independence, documentation, non-PII data extraction, privacy and applicable laws, standards and best practices, and do-no-harm-by-design. Applicants submit evidence online and a technical team reviews against each indicator. Approved status lapses after a year unless the tool is reassessed.

IEEE CertifAIEd, applied by the City of Vienna. The programme assesses an AI system against ethics criteria covering transparency, accountability, algorithmic bias, and privacy, and issues a mark on the result. Vienna took the mark on a municipal service, so a city government rather than a vendor carried it.

High for the registry pattern; moderate for the CE-marking analogue. The DPGA model (open standard, evidence-based review, public registry) is directly transferable to a national registry of verified civic technology. FedRAMP's do-once-use-many model solves a real coordination problem, but its cost structure has to be reworked for small builders.

The EU's digital CE marking is the precedent that matters most for agents: a machine-readable trust signal embedded in the product interface rather than a registry listing alone, which is the form an agent can query programmatically.