"Should robots pay taxes?" As ridiculous as this question may sound at first glance, it is at the core of the debate that we should have about the future of the tax and benefit systems as the arrival of Artificial Intelligence might necessitate a radical rethinking to safeguard the sustainability of public finances. 

The rapidly increasing application of the AI technologies will undoubtedly impact how we work in the future. On the one hand, the new technologies have the potential to improve workers' lives by – among others – decreasing the relative amount of onerous or tedious work and by facilitating labour market access for the elderly and handicapped or those who wish to work flexibly and/or remotely. On the other hand, AI also clearly brings with itself the risk of substituting human labour by machines on a scale not witnessed before as it is clear that the propagation of AI will destroy some jobs, transform others and hopefully create new ones. The jury is still out on the net impact that AI will have on the number of jobs. Experts argue about the quantification of AI's effect on jobs and their estimates range from an overall negligible impact[1] to a loss of 47%[2] and 54%[3] of jobs in gross terms in the US and Europe.

Based on the experience of the past episodes of rapid technological progress, one can presume that in the long term workers will reallocate between tasks, companies or sectors and that the augmentation effect (whereby humans work alongside machines) will prevail. However, in the short- and medium-term the displacement of workers clearly poses challenges as the reallocation will not happen instantaneously and the disruption along the way is bound to put strain on the labour market and as a result also on public finances.

The displacement of workers is very likely to put significant pressure on the revenue side of public budgets, if the current system based on personal income tax and social contributions does not adapt. This impact on tax revenues might be additionally exacerbated by the shift from employment based on formal open-ended contracts (with lower administrative burden since tax collection is facilitated by employers) to freelance/gig work (more complex to tax).  It is likely to also affect public expenditure by increasing spending on unemployment benefits, active labour market policies and education.

There is an emerging consensus in the literature that medium-skilled workers are likely to lose out disproportionately in comparison to low- and high-skilled workers as their tasks lend themselves more easily to substitution by AI machines (a phenomenon described as the polarisation of the labour market). Whereas earlier it was possible to substitute human labour with technology only for the routine part of tasks, the proliferation of AI is increasing the potential for the automation of tasks requiring those cognitive skills that can be captured in an algorithm. At present this concerns predominantly medium-skilled workers (for example it is easier for a computer to replace a bank teller than a cleaner). This risks increasing the growing income inequality, which is a major policy challenge when aiming at making growth more inclusive. The shrinking labour income might constrain Member States' abilities to redistribute through labour taxation, further exacerbating risks of increasing income inequality. Shifting to tax bases such as capital income or stock could be explored to ensure both sufficient revenues and a fair burden sharing. These considerations are reflected in the discussion dubbed as "taxing robots".

The shift to robotised work does not only create challenges for taxation, but also for the education of the labour force. In that context, ensuring equal access to quality education should be another way to mitigate AI's inequality-inducing effect. It is essential to make sure that there are enough experts to develop and implement AI. But equally importantly is for the educational systems to equip the population both with at least basic digital skills and the ability to think critically and creatively in order to make the most of the opportunities offered. Given the increased importance of the education level for workers to succeed in the era of AI and the potential reluctance on the part of employers to enhance skills of workers who become more mobile, the state might have to pick up the tab to both upgrade skills of existing workers but also to adjust educational systems so that they support development and uptake of human-centric AI.

Even though there is uncertainty about the exact impact the AI's diffusion will have on the current economic systems, policy-makers should be ready with their policy response in order to maximise the benefits that the new technologies bring but at the same time mitigate the negative effects. In order to raise awareness among colleagues of AI-related issues from the economic point of view, DG ECFIN recently organised an internal seminar during which Mario Mariniello, the Digital Advisor to the European Political Strategy Centre, presented his paper on "The Age of Artificial Intelligence: Towards a European Strategy for Human-Centric Machines". This provided an opportunity for colleagues to discuss AI's impact on the labour market, public finances, productivity, Europe's place in the AI race but also broader perspective issues such as the ethically-sound approach to AI.   

[1] Gordon, Robert J. (2015). "Secular Stagnation: A Supply-Side View." American Economic Review

[2] Frey, C. B., & Osborne, M. (2013). The future of employment. How susceptible are jobs to computerisation. Working Paper. University of Oxford

[3] Bowles, Jeremy. (2014) ''Chart of the Week: 54% of EU jobs at risk of computerisation''. Blog Post, Bruegel. July 24, 2015. Retrieved from: http://bruegel.org/2014/07/chart-of-the-week-54-of-eu-jobs-at-risk-of-c…