From Malta to Poland: each EU state to have AI strategy

"Like the steam engine or electricity in the past, AI is transforming the world," it added.

"In order to maximise investments, pool important resources such as data, provide a seamless regulatory environment, all member states need to put in place national AI strategies," the paper said.

The document was drafted by the European Commission, but the text went back and forth between Brussels and national capitals several times – the member states have effectively signed off on its content.

Currently, only Germany, Finland, France, Sweden, and the UK have specific national AI strategies – and the UK will soon leave the EU.

Of the 23 remaining EU states without such a strategy, 13 are working on one. A few others have a broader digitalisation strategy in place that deals partly with AI.

The national plans should address how much the countries will invest in AI, and how the member states will improve their citizens' digital skills.

"Poor general technical knowledge in the broader population hampers the accessibility and uptake of AI-based solutions," the paper warned.

However, it noted that it will be up to member states to decide on the "exact form, contents and governance" of the national strategies.

The document itself is non-binding and there are no sanctions if a member state ends up without an AI plan.

Indeed, many of the actions to be taken at national level are phrased in a rather non-committal way, saying for example that all states are "encouraged" to publish their national strategies by mid-2019 – and are "encouraged" to create legal environments where new AI applications like autonomous driving can be tested.

The cautious language was asked for by some member states, according to a source in the European Commission who spoke on the condition of anonymity.

"It was difficult for some to have strong commitments before their national debate is done," the source told EUobserver.

Conversely, the EU strategy paper also serves a domestic political purpose, the EU contact explained.

The paper was drafted in cooperation with member states' economy and industry ministries. These can now point to the EU paper as an incentive to convince their national peers of the need to invest in AI.

'Ethical' AI

Friday's 'coordinated plan' is published less than eight months after the commission's own strategy paper on AI, and much of the direction is still the same.

The new paper stressed again that Europe should be a "world leader in ethical, trusted AI", with a human-centric approach.

It repeated the goal that Europe should spend at least €20bn in the period 2018-2020, followed by the same amount per year in the period 2021-2030.

The paper also said that the EU will set up in 2020 a database of anonymised images from cancer patients - in the hope of improving diagnosis, treatment and follow-up of the disease.

The document also noted that the EU will check in 2019 whether the regulatory framework is "fit for purpose" for AI-enabled technologies - in particular for connected and automated driving.

Later this month, a group of AI experts tasked by the commission will present their draft ethical guidelines on AI - to be followed by a final version in March.

In the first half of 2019 the EU will then organise an international ministerial meeting "with the aim of forging a global consensus on the ethical implications of AI".

Self-driving cars for example will need to be programmed in such a way that the car can make an autonomous decision when faced with the dilemma of killing the driver or killing a pedestrian.

Other ethical questions on AI involve the data that goes in the system. If that data is already biased, then AI will not take fair decisions.

There are well-known examples of 'crime-predicting' AIs that had racist tendencies, because the available data on crime was gathered with an unfair focus on a certain ethnic group.

The document presented on Friday itself is supposed to be flexible, and will be kept up to date annually.