There is a reason why the AI Act proposal is complemented with the reviewed Coordinated Plan on AI. You can’t just establish a legal framework on AI, cross your fingers and see it works. You need something more for the borderless stuff in the borderless world. The new Coordinated Plan with Member States is exactly the establishment of a global level playing field for the technology that transcends borders.
The EU is still about the Union level and the Member States level. While the key components of successful implementation of an AI approach are dispersed across both levels. And whatever the European Commission, the European Parliament, and the European Council do at the Union level, it is just not enough because some issues are beyond their competence having intentionally left for the Member States to deal with as those see it fits. The coordination allows exactly what’s missing here in order to join efforts.
The Coordinated Plan on AI in focus is the 2021 review of one adopted back in 2018. It is aimed at strengthening leading position globally while fostering development, consolidating investments, and removing fragmentations. Its framework is as follows.
- governance & coordination framework
- computation infrastructure
- stakeholders collaboration
- research capacities
- take-up environment
- scale-up support
For People For Good
- talents & skills nurturing
- trust ensuring policy
- global promotion
- public sector
- law enforcement plus
The 2021 Coordinated Plan on AI reiterates what’s important, provides overview of actions already taken, outlines prospects for the future. With the novel legal framework on AI (a.k.a. AI Act) in mind. National AI strategies have their special place within the governance & coordination framework subset of the Enabling Conditions set. And different kinds of means, vehicles, instruments are detailed across all the subsets—what the European Commission will do, what the Members States are encouraged to do, what they will do together.
So, what the global promotion stands for? Having buried deep inside, that is far from being of secondary importance. What we see here is actions not sweet dreams. An approach to frame a digital phenomenon in the real world.
The international fora—UN, UNESCO, OECD, the Council of Europe, the G7 and the G20—and international enablers—ISO, IEEE, WIPO—are not left behind. But what really matters is cooperating with ones while preventing others from having the essentials for AI-enabled technologies to be operational.
Here’s how exactly the EU is active at the Union level:
- a joint EU-Japan committee on AI held its first meeting in November 2020
- options for reinforced cooperation with Canada on AI have been discussed
- work on a joint AI task force with India has started
- there are plans to start discussions with Australia and Singapore
- setting up an EU-US Trade and Technology council in its proposal phase and AI Agreement with the US is another target
- the Dual-Use Regulation is being revised with specific purpose to ensure that no AI from within the EU is misused
Those bilateral efforts will be increased and new ones will inevitably emerge. But it is the Dual-Use Regulation dimension what deserves special attention here.
The revision indeed originates from the 2016 proposal of the European Commission and it is still in progress. But the European Commission, the European Parliament, the European Council has reached agreement on the final text, the European Parliament has already adopted that, and the European Council is expected to agree to (if has not already done that). It is just 90 days after publication then for the revised Dual-Use Regulation to come into effect. So, why does it matter in terms of preventing from AI being misused?
Notwithstanding the pre-AI age language, the ‘cyber-surveillance items’ defined as dual-use items designed to enable the covert surveillance of natural persons by monitoring, extracting, collecting or analysing data from information and telecommunication systems. It is further set that an authorisation shall be required for the export of even not listed cyber-surveillance items if the exporter has been informed that those in question are or may be intended, in their entirety or in part, for use in connection with internal repression, with the commission of serious violations of human rights and international humanitarian law, and where the exporter is aware of that according to its due diligence findings, the exporter shall make it notified while the competent authority shall decide whether or not to make the export concerned subject to authorisation. So, it does matter.
Another reason why the Dual-Use Regulation matters is that other jurisdictions beyond the EU are guided by that. Take Ukraine. It was 2018 when the Cabinet of Ministers of Ukraine adopted Unified List of Dual-Use Items mirroring the Dual-Use Regulation list in order to bring Ukrainian regulations in this area in compliance with the EU standards. It is not clear now whether or not Ukraine is to reimplement the revised Dual-Use Regulation at all and whether or not the unlisted cyber-surveillance items are to be covered. But how can we be sure about it? In no way can we deny that companies planning to export AI-enabled technologies will be regarded as exporting the cyber-surveillance items and will be required to obtain prior authorisation by the State Export Control Service of Ukraine.
Let us now step back and see the bigger picture. If even a dual-use regulation covers AI-enabled technologies and gets as far as Ukraine, are there any doubts that the EU is going to use anything at its disposal in order to impact AI globally?