|
Key extracts from the report on the public consultation
In total, 1215 contributions were received, of which 352 were on behalf of a company or business organisations / associations, 406 from citizens (92% EU citizens), 152 on behalf of academic / research institutions, and 73 from public authorities. Civil society voices were represented by 160 respondents (among which 9 consumer’s organisations, 129 non-governmental organisations and 22 trade unions). 72 respondents contributed as “others”.
Of the 352 business and industry representatives, 222 were companies and business representatives, 41.5% of which were micro, small and medium-sized enterprises. The rest were business associations. Overall, 84% of business and industry replies came from the EU-27.
Mandatory requirements
42% of respondents request the introduction of a new regulatory framework on AI, another 33% think that the current legislation needs to be modified in order to address the gaps identified. Only 3% think that current legislation is fully sufficient.
Respondents seemed to agree with all the mandatory requirements proposed by the White Paper with high percentages ranging from 83% to 91% for each requirement. Clear liability and safety rules (91%), information on the nature and purpose of an AI system (89%), robustness, and accuracy of AI systems (89%). Human oversight (85%), quality of training datasets (84%) and the keeping of records and data (83%).
Concerning the scope of this new possible legislation, opinions are less straightforward. While 42.5% agreed that the introduction of new compulsory requirements should only be limited to high-risk AI applications, another 30.6% doubt such limitation. It is interesting to note that respondents from industry and business were more likely to agree with limiting new compulsory requirements to high-risk applications with a percentage of 54.6%.
However, several respondents do not appear to have a clear opinion regarding what high-risk means: although 59% of respondents support the definition of high-risk provided by the White Paper, only 449 out of 1215 (37% of consultation participants) responded to this question.
Biometric identification
Respondents had doubts on the public use of [biometric identification systems] with 28% of them supporting a general ban of this technology in public spaces, while another 29.2% required a specific EU guideline or legislation before such systems may be used in public spaces.
Finally, 6.2% of respondents did not think that any further guidelines or regulations are needed.
Enforcement and voluntary labelling
To make sure that AI is trustworthy, secure and in respect of European values, the White Paper suggests a series of conformity assessment mechanisms for high-risk applications. Of those mechanisms, 62% of respondents supported a combination of ex-post and ex-ante market surveillance systems. 28% support external conformity assessment of high-risk applications. 21% of respondents support ex-ante self-assessment.
Voluntary labelling systems could be used for AI applications that are not considered of high-risk. The 50.5% of respondents find it useful or very useful, while another 34% do not agree with it. 15.5% of respondents declared that they do not have an opinion on the matter.
Safety and liability implications of AI, IoT and robotics
60.7% of respondents supported a revision of the existing Product Liability directive to cover particular risks engendered by certain AI applications.
Among the particular AI related risks to be covered, respondents prioritised cyber risks with 78% and personal security risks with 77%. Mental health risks follow with 48% of respondents flagging them, and risks related to the loss of connectivity, flagged by 40% of respondents.
Moreover, 70% of participants supported that the safety legislative framework should consider a risk assessment procedure for products subject to important changes during their lifetime.
|
