Pilots
Time-limited evaluations of AI systems.
How the laws define this
Regulators uniformly characterize pilots as time-limited trials, evaluations, or experiments designed to test technologies under controlled conditions and assess their technical and legal feasibility before scaling. Each framework requires formal evaluation metrics, such as success metrics, key performance indicators, or post-pilot reviews, to determine whether a solution is ready for scale-up. However, these instruments diverge notably in their underlying domain, operational settings, and specific assessment criteria. The Digital Taskforce for Artificial Intelligence focuses specifically on assessing the organizational readiness and legal aspects of artificial intelligence solutions, explicitly highlighting data protection impact assessments for personal data. In contrast, the Smart City Blueprint for Hong Kong 2.0 concentrates on trialing technologies and service models in controlled public settings to ensure compliance with broader sectoral standards. Differing from both, the National Innovation Centre for Privacy Enhancing Technologies restricts pilots to privacy enhancing technologies, specifically measuring privacy-utility trade-offs, testing system architectures, and validating governance arrangements for defined use cases.
Synthesised from the 3 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| Denmark | Digital Taskforce for Artificial Intelligence | In Force | Adds defined success metrics, legal assessments like DPIAs for personal data, and iterative learning to inform scaling decisions. |
| Hong Kong | Smart City Blueprint for Hong Kong 2.0 | In Force | Specifies testing in controlled public settings with pilot-stage assessments, KPIs, post-pilot reviews, and compliance with sectoral standards. |
| Netherlands | National Innovation Centre for Privacy Enhancing Technologies (NICPET) | In Force | Limits pilots to PETs implementations or experiments validating technical feasibility, legal admissibility, governance, privacy-utility trade-offs. |
Definitions (3)
Time-limited evaluations and testbed activities intended to assess technical, legal and organisational readiness of AI solutions under controlled conditions; pilots include defined success metrics, legal assessments (e.g., DPIAs where personal data is involved) and iterative learning to inform scaling decisions.
Time-limited trials intended to test technologies and service models in controlled public settings, subject to pilot-stage assessments, KPIs, post-pilot reviews and evaluation to determine scalability and compliance with sectoral standards.
A time-limited implementation or experiment using PETs to validate technical feasibility, legal admissibility and governance arrangements in a defined use case. Pilots are used to test architectures, measure privacy-utility trade-offs, and produce artefacts for potential scale-up.
Not to be confused with
AI Pilots specifically denote controlled, time-bound projects for AI systems, while these definitions apply pilots more broadly to AI solutions, technologies, or PETs.
AI sandbox/pilot emphasizes a controlled environment for time-limited AI experimentation, whereas these definitions focus on trials, testbeds, or implementations with assessments.
Testbeds/sandboxes are controlled environments for AI experimentation, but these pilots additionally include defined metrics, legal reviews, KPIs, or privacy trade-offs.
Questions and answers
What legal assessments are part of pilots under Danish AI policy?
Pilots include legal assessments such as DPIAs where personal data is involved. They also feature defined success metrics and iterative learning.
What evaluations follow pilots in Hong Kong's Smart City Blueprint?
Pilots undergo pilot-stage assessments, KPIs, post-pilot reviews, and evaluation for scalability and sectoral standards compliance.
How does NICPET use pilots for privacy technologies?
Pilots test PETs architectures, measure privacy-utility trade-offs, and produce artefacts for scale-up in defined use cases.
Related Terms
AI Pilots
Controlled, time-bound projects to test AI systems....
AI sandbox / pilot
Controlled environment for time-limited AI experimentation....
Pilot
Time-limited experiment implementing PETs....
Pilot Project
Limited-scope, time-limited technical implementation co-designed under the MoU....
testbeds/sandboxes
Controlled environments for AI experimentation....