EP Input on AI
In a resolution of 20 January 2021, the European Parliament outlined definitions and several ethical principles as regards the application of artificial intelligence (AI) in the military and civilian sectors. There is no focus on tje use of AI in criminal justice, however the resolution includes a number of statements on the use of AI for justice in general as well as on the challenges for big data analyses, e.g., facial recognition.
MEPs emphasised that in any area, especially those managed by the state such as justice, AI must remain a tool used only to assist decision-making or help when taking action. AI must be subject to human control, allowing humans to correct or disable it in case of unforeseen behaviour. Moreover, AI is a scientific advance which should not undermine the law but should, on the contrary, always be governed by it. Under no circumstances should AI, robotics and related technologies violate fundamental rights, democracy and the rule of law.
Regarding the use of AI in the field of justice, MEPs adopted the following positions:
- The use of AI in fighting crime and cybercrime could bring a wide range of possibilities and opportunities but, at the same time, the principle “what is illegal offline is illegal online” should continue to prevail;
- The option of whether it is appropriate for law enforcement decisions to be partially delegated to AI should be discussed;
- When using evidence provided by AI-assisted technologies, the judicial authorities should have the obligation to provide reasons for their decisions;
- Research should explore improvements in the analysis and collection of data and the protection of victims, whereby it must be ensured that safeguards for due process and against bias and discrimination, with the precautionary principle are being applied;
- The principles of governance, transparency, impartiality, accountability, fairness and intellectual integrity in the use of AI in criminal justice are important;
- It must be guaranteed that the pubic is kept informed about the use of AI and that decisions are personally taken by the responsible official who can, if necessary, deviate from the results received from AI;
- Defendants must have the right to appeal to appeal the decision; such an appeal should be decided without the use of an AI system.
The resolution also addresses the phenomenon that AI technologies can have a deep impact on fundamental rights. The Commission is called on to assess the consequences of a moratorium on the use of facial recognition systems until the technical standards can be considered fully fundamental rights-compliant and that there are strict safeguards against misuse. Mass scoring applications (monitoring and rating citizens) should be explicitly banned by public authorities. The EU should better promote the EU’s viewpoint in these areas when negotiating laws at the international level. The resolution also concerns over “deepfake technologies” that have the potential to “destabilise countries, spread disinformation and influence elections”. Creators should be obliged to label deepfake material or any other realistically made synthetic videos as “not original” and more research should be done into technology to counter this phenomenon.
The EP resolution is considered giving an input into the Commission’s White Paper on AI presented in 2020 (→ eucrim 1/2020, 8-9). The Commission is expected to present a legislative proposal on AI in spring of 2021.
Independently of the EP resolution, a virtual conference on AI and human rights was held on 20 January 2021, under the German presidency of the Council of Europe. The basic demand of this conference was to create an international legal framework for AI at the level of the Council of Europe. This could consist of both mandatory and soft law components, but always with respect to human rights, democracy and the rule of law.