The goal of the approved internal protocol “Definition of work methodologies and protocols for the implementation of algorithmic systems” is to ensure that any artificial intelligence (AI) systems developed at a municipal level are used in accordance with legal, ethical and technical standards. The protocol specifically defines the means for protecting citizens’ rights in the process to tender and implement such systems.
The protocol is a pioneering document in the regulation and governance of AI at a local level. It combines the public procurement process, the life cycle of algorithmic systems, existing data protection mechanisms at the City Council and the ethical standards defined by various administrations and organisations.
The protocol is based on the proposal by the European Commission to regulate AI, establishing different types of action to guarantee rights depending on the level of risk from algorithmic systems. Unacceptable risks are those which represent a clear threat to the security of people’s rights and are rejected automatically. For high-risk algorithmic systems with a direct impact on people’s basic rights, the protocol sets out a series of stricter guarantee mechanisms than for systems regarded as posing a limited or minimal risk, such as studies with an algorithmic impact developed for an external advisory body or the creation of communication and transparency mechanisms, such as algorithm registers.
Activation of the regulatory framework
The protocol comes within the framework of the “Government Measure on the Municipal Algorithm and Data Strategy for the Ethical Implementation of Artificial Intelligence”, approved in April 2021. The measures set out twenty actions for rolling out artificial intelligence in the various municipal services. The sixth of these referred to the definition of work methodologies and protocols, now being developed through this internal protocol.
Other actions contemplated by the Government Measure will be activated in 2023. Specifically, these include the creation of a public register of algorithms and an external advisory body whose tasks will include the compilation of algorithmic impact studies defined in the document.