For Caribbean and international professionals working on government digitalisation

Logo Rijksoverheid, to the homepage

3.3 Regulate Algorithms

Social challenges

It is not sufficiently clear to society which standards based on public values should be guaranteed for algorithms. It is unclear when algorithms are used, even if they directly impact users. Specifically for the government, it should be clear to citizens when an algorithm has been used in decision-making, especially when it has the potential to impact their situation. It is also important that these algorithms are not perceived as a “black box”. The government must set a good example by experimenting with values-driven AI applications that support societal tasks.

Results achieved by 2023

  • Efforts towards the EU’s AI Act have resulted in human rights featuring more prominently in risk assessments for high-risk AI systems and supplementary measures that strengthen transparency and the rights of natural persons.
  • An algorithm registry has been launched.
  • A vision for generative AI is in the works and will be published in 2024.
  • An initial version of the “Use of algorithms” algorithm implementation framework (algorithm framework) provides insight into the key standards.
  • ELSA labs (ELSA stands for ethics, legal and societal aspects) for people-oriented AI have been launched, in which participants collaborate to produce algorithms that satisfy human rights and public values.
  • An AI Validation Team has been established within (Ministry of Interior and Kingdom Affairs) BZK in which software engineers and policymakers work together to gain knowledge and experience in validating algorithms. In addition to creating tools, they make the risks and opportunities of generative AI measurable and explore what datasets are needed to test AI applications.

Goals & indicators

Goals and indicators
Goals Indicators
1. We set clear requirements for the use of algorithms by government organisations to arrive at responsible and innovative generative AI applications in government. This includes the creation of a uniform algorithm framework— supported by government organisations—for the deployment of algorithms. Requirements cover topics such as roles and responsibilities regarding the review of algorithms and AI (governance), the application of a uniformly developed methodology to detect bias/discrimination, the implementation of human rights tests (e.g., IAMAs), and generally applicable procurement conditions for algorithms that the government sources from third parties.
  • There is clarity on the application of requirements on the use of algorithms within government organisations.
  • There is a clear overview of tools and resources to help governments meet legal standards.
2. The government is transparent about the use of algorithms.
  • A well-stocked government algorithm register that provides insight into high-risk and high-impact algorithms in use.
  • New high-risk algorithms used by the central government are included in the algorithm register and subjected to a human rights test.
  • Departments report annually to the House of Representatives on the progress of including algorithms in the register.
3. We will ensure further improvement of AI and algorithm surveillance, targeting both companies and governments.
  • Availability of explanation of standards and guidance on regulations applicable to AI and algorithms.
  • Improve algorithm risk signalling function; reports, research into service desk function for citizens and businesses.
  • Increase cooperation among digital regulators (including colleges, inspectorates and market regulators).
  • Establish surveillance tasks under the AI Regulation.
4. We set requirements at the European level for technology, such as generative AI systems, to ensure that they are secure and in line with our public values.
  • In the negotiations on the European AI regulation, we are actively negotiating the requirements for high-risk AI systems that protect fundamental rights and preserve AI innovation opportunities.
  • We encourage government organisations offering high-risk AI systems in the Netherlands to comply with the requirements of the AI Regulation.
  • We conduct practical research on the implementation of the AI Regulation in government with a focus on open standards and their connection to existing laws and regulations.
  • The Council of Europe treaty must ensure that the application of AI is in line with human rights, democracy and the rule of law. Together with the European Commission and member states, we are moving to ensure that global players commit to this treaty.

What are our forthcoming actions?

To find out the goals we are setting for the upcoming year to further regulate algorithms, see priority 3.3 actions.

 

Got a query, thought, comment, or suggestion?

If you're working on digitalising the government and got something on your mind, please share your thoughts with us.