The concept of "common European data spaces" was endorsed in a new coordinated plan on AI developed by the Commission in partnership with the 28 EU member states, Norway and Switzerland.
According to the plan, each country and the Commission would identify and share public data sets that could be made accessible to businesses and public sector bodies to inform the development of AI tools.
The Commission proposed that, beyond 2020, the EU, member states and businesses could co-invest €1 billion "in the creation of a common European Data Space that makes data easily available for re-use to innovators, businesses and public sector".
The plan also envisaged a role for "next-generation digital industrial platforms", and said new data infrastructure could also be developed to enable "real-time" sharing of data and the use of a "data sandbox" for testing "data-driven AI-powered services".
Two major data and AI projects were proposed for the health sector.
The first project involves linking genomics repositories across the EU, and helping to build rare disease registries too.
In the second project, the Commission said it will work with EU member states to develop a "common database of health images initially dedicated to the most common forms of cancer (anonymised, and based on data donorship by patients)" in 2020.
The Commission called on the European Data Protection Board to develop guidelines on the processing of personal data in the context of health research. It said the guidance "will facilitate the development of large cross-country research datasets that can be used for AI".
"AI needs vast amounts of data to be developed," the Commission said in its communication published alongside the coordinated plan. "Machine learning, a type of AI, works by identifying patterns in available data and then applying the knowledge to new data. The larger a data set, the better AI can learn and discover even subtle relations in the data.
"Once trained, algorithms can correctly classify objects that they have never seen, in more and more cases with accuracies that exceed those of humans. Hence, access to data is a key ingredient for a competitive AI landscape, which the EU should facilitate in full respect of personal data protection rules," it said.