General-Purpose AI fit for European small-scale innovators

Big Tech wants to exclude General-Purpose AI from the scope of the AI Act. In response, the European DIGITAL SME Alliance, representing a network of 45.000 ICT small and medium enterprises in Europe, calls for balanced obligations across the AI value chain to support SMEs, protect fundamental rights and advance digital sovereignty.

One of the key ambitions of the EU’s Artificial Intelligence Act is to foster innovation. We believe that much more can be done in this domain. A significant number of AI developers, including European SMEs, are dependent on ‘AI as a service’ which is provided by a handful of companies. This is because small-scale innovators lack the financial and technical resources to train the underlying models for cutting-edge AI systems. Therefore, start-ups and SMEs tend to access large pre-trained models through APIs. In this course, many small and medium enterprises have found commercial success.

This success is now under threat. If the AI Act places obligations for such systems on downstream users, SMEs will be overburdened with high compliance costs. At best, this will prove to be overly burdensome for SMEs. Only a handful of providers have the resources to develop cutting-edge AI, while potentially hundreds of smaller companies will build their business models on API access to such AI. Ensuring that these systems are adhering to the AI Act’s requirements before they are sold will reduce the costs of compliance for SMEs and start-ups.

At worst, it will be impossible for SMEs to comply with the requirements laid down in the AI Act. For example, they will not have access to the source data (Art 10), nor be able to conduct full technical documentation (Art 11) or ensure accuracy and robustness is built into the system (Art 15). Moreover, it will be impossible for SMEs to guarantee that upstream General-Purpose AI (GPAIs) have been developed under satisfactory Risk Management system (Art 9) and Quality Management system (Art 17). These processes would inflict heavy costs and would also make SMEs entirely dependent on the good faith of upstream GPAI developers to guarantee these provisions.

In both above-mentioned scenarios, the competitiveness of the market is at risk. Because the handful of companies that develop GPAI are not based in Europe, this will also significantly affect Europe’s efforts towards digital sovereignty.

We therefore propose that providers of cutting-edge general purpose AI systems, such as large language models, carry a fair share of the burden, and ensure their systems are made to conform to European requirements as much as possible before being placed on the market. By doing so, smaller players will have reduced compliance costs, and therefore lower barriers to entry in the market, which will boost European innovation.

Balancing obligations in the AI value chain is one of the key improvements that DIGITAL SME has asked for in the AI Open Letter. The Letter has been signed by over 50 small and medium sized innovators and contains specific proposals on how to make the AI Act work for everyone.

Finally, we want to use this occasion to raise awareness on the covert misleading ways that have been used to influence lawmakers on the AI Act. In July, a letter sent by three Members of the European Parliament to its President, Roberta Metsola highlighted that Big Tech companies use shell organisations to mask themselves as representatives of start-ups and SMEs while promoting the positions of their funders. Recently, there has been noticeable effort from Big Tech lobbying groups to exclude General Purpose AI from the scope of the AI Act. Members of the European Parliament should take note of these methods when assessing their policy options.