Only the US can do security cooperation with NATO on AI---EU is far behind on funding, adoption, and private investment.
Christie 22, Senior Research Fellow for the Finnish Institute of International Affairs, Research in the areas of International Affairs and International Security, with a particular focus on: Geoeconomics, Economic Statecraft, Defence Economics, Defence Innovation, Artificial Intelligence, NATO, Great Power Competition., (Edward Hunter, “Defence cooperation in artificial intelligence: Bridging the transatlantic gap for a stronger Europe” in European View, Volume 21(1), Wilfried Martens Centre for European Studies, https://journals.sagepub.com/doi/pdf/10.1177/17816858221089372) //CHC-DS
Before proceeding, it is worth spelling out the extent to which European security is dependent on NATO and in particular on the US. Of the EU’s 27 member states, 21 are members of NATO. These countries account for about 93% of the population1 of the EU. Within NATO, those Allies that are also EU members only account for about 20% of total defence expenditure across the Alliance, while the US alone accounts for about 70% of the same total.2 Beyond these aggregate indicators, it is furthermore the case that the US is considerably ahead of the EU in terms of practical adoption of AI. For illustration, in 2020 US private-sector investment in AI was around $23.6 billion, but was only $2 billion in the EU, implying a ratio of 12 to 1 in favour of the US (Zhang et al. 2021, 96). Scientific output indicators offer a more nuanced picture. In 2019, the EU accounted for 16.4% of the world’s peer-reviewed AI publications, ahead of the US with 14.6%, while China occupied the top spot with 22.4% (Zhang et al. 2021, 20). On the other hand, if one measures research output in terms of publications on the Arxiv database, the US is ahead of the EU (Zhang et al. 2021, 33) by a ratio of almost two to one, which is nonetheless much less than the large gap in private investment mentioned above. That the EU performs similarly to the US in terms of scientific research, but far less well in terms of investment and commercialisation of new digital technologies, is an old problem which has proven very difficult to address, whether at national or EU level (Baroudy et al. 2020).