The EU Parliament discussed Wednesday (May 11) a draft report with the proposed amendments to the Artificial Intelligence Act, and while the different parties seem to agree on the most important issues in the regulation, lawmakers will still have to iron out a number of differences before the text is put to a vote later this year.
The two co-rapporteurs, who lead the efforts to get an agreement at the Committee level before sending the text for a vote in the plenary, agree with the risk-based approach adopted by the Commission. That is, the obligations set out in the AI Act only apply to forbidden practices, to high-risk AI systems, and to certain AI systems that require transparency. But, against the request made by some lawmakers, no AI system is yet excluded ex-ante, either from the definition of “artificial intelligence” or by carving out exceptions for particular types of AI systems, including general-purpose AI. Some lawmakers are still suggesting that general-purpose AI should be excluded from the regulation, and it is possible that more amendments will be added to the text in this regard.
The lawmakers also agreed on adding a number of cases to the list of high-risk AI systems, which are subject to more oversight and prohibitions. For instance, AI systems used to influence or shape the development of children should be considered high risk. AI systems used by candidates or parties to influence voters in local, national or European elections, and AI systems used to count such votes, will also be considered high risk. Deepfakes and editorial content written by AI will also be added to this list. Lawmakers stressed that high-risk AI systems are not prohibited per se, nor are they to be seen as undesirable, but they may be subject to limitations and prohibitions under certain circumstances.
Some lawmakers requested to go one step further and impose an outright ban on certain AI systems, like biometric AI systems in public spaces or for armed forces. Yet, one of the rapporteurs warned that complete bans could curtail innovation. It is likely that more amendments will be added requesting specific prohibitions to certain AI systems.
The report also proposes to rebalance the responsibility among the actors involved in the AI value chain based on the role they play in the design or implementation of the AI system. In this type of system, sometimes a provider may not have access to the data or an algorithm, but a user may be in a better position to detect when data has been wrongly used, a rapporteur argued. Thus, the responsibilities for data governance will be allocated accordingly.
The parliament also proposed to give the European Commission the role of European enforcer when certain AI practices are widespread across the region. The EU regulator will be given powers of market surveillance authority as well as the right to impose hefty fines. For high-risk activities, the parliament expects the Commission to be a tough enforcer.
The EU lawmakers will now have until June 1 to table new amendments to the text. The Joint committee will meet again on June 30 to discuss the proposed amendments. The draft report will be put to a vote by the two committees in late October, and in plenary in November.
The AI Act, which is considered a main pillar of the EU digital single market strategy, will set horizontal rules for the development, modification and use of AI-driven products, services and systems within the territory of the EU, applicable to all industries.
It aims to ensure the proper functioning of the internal market and the free movement of AI-based goods and services cross-border and to codify the standards of the EU trustworthy AI paradigm, which requires AI to be legally, ethically and technically robust, while respecting democratic values and fundamental rights.
Read More: EU Parliament Wants the EU to Be an AI Global Standard-Setter
We’re always on the lookout for opportunities to partner with innovators and disruptors.
Learn More