Senate defines rules for AI use, but offices may choose not to join
The Federal Senate published a set of rules that guide the use and development of Artificial Intelligence Systems (AI) within the administrative structure of the house.
The document, signed by Director-General Ilana Trombka, has immediate application to secretariats, commissions and internal sectors, but does not automatically extend to parliamentary offices, which may choose to join or not to the regulation.
The initiative places the Senate among the first organs of the Legislative Power to establish its own guidelines for AI use, amid the advancement of automated tools at different spheres of the public service.
Motta gives progress on requests for cassation against Eduardo Bolsonaro; See arguments
Movement occurs after PT and PSOL have funds against the parliamentarian, which has been in the United States since March this year
Lula wants to block networks for children under 12 and impose parental control up to 16
PL draft foresees age verification, prohibition of harmful ads and punishments to platforms
Central principles
The normative states that AI systems must respect fundamental rights, promote social welfare, ensure human supervision in all stages and protect personal data and legal confidentials. The text still predicts:
• Use of reliable, auditable and preferably public data;
• Transparency on the impacts, audits and tool monitoring;
• Mandatory human validation of all automated decisions;
• Continuous users’ training;
• Measures against algorithmic discrimination and abusive bias.
AI use should be done in an “ethical and responsible” manner, with an express prohibition of the use of protected or personal data without consent, in addition to the requirement of explicit indication when automated systems are involved in administrative or analytical proceedings.
The text also details the technical criteria for the internal development of Senate AI systems. Every project should follow good data governance practices, with attention to the quality, integrity and security of the information used.
The development cycle should be documented clearly, including:
• Data selection and processing;
• Justification for the models used;
• Training and validation techniques;
• Identified risks and mitigation strategies.
The guidance is that any initiative involving would be traceable, auditable and follow the standards set by national and international security and data protection standards.
