Revising the AI White Paper of the EU Before Its Publication: Key Points to Consider
The European Commission is set to release a white paper aimed at supporting the development and uptake of artificial intelligence (AI), with the initial public version expected in mid-February. This move comes as the Commission grapples with the need to regulate AI while avoiding stifling innovation and competition.
One of the key concerns is the potential exclusion of foreign companies from the European market. Requiring companies to use only EU-approved datasets for training AI systems could reduce competition and options for consumers and businesses. This could be particularly problematic for high-risk sectors, where the stringent regulations proposed in the EU AI Act, such as transparency obligations, mandatory documentation, audits, and heavy fines, may slow AI development and deployment.
Another concern is the risk of intellectual property loss during reviews, which might deter companies from launching AI products and services in Europe. The Commission recognizes this risk and is encouraged to focus on regulating specific activities related to AI, such as fair hiring practices, rather than creating national rules that could disrupt the digital single market.
The Commission also faces the challenge of ensuring that European data is representative and diverse enough for developing systems deployed globally. Using only EU-approved datasets could put consumers at risk by limiting the diversity of AI systems.
To address these concerns, the Commission should review the feedback from companies participating in the High-Level Expert Group's piloting phase to ensure any rules it does create take into account the specificity and diversity of AI systems. It should also encourage the continued development and testing of voluntary industry best practices for AI.
The Commission should also be mindful of the potential bureaucratic hurdles and compliance costs created by the regulations. Singling out entire sectors as high-risk and covering them with sweeping rules could limit the deployment of AI in these sectors, even if they have low-risk AI applications. Ex-ante conformity assessments for a broad set of AI products and services before being allowed on the market could make it more expensive and time-consuming for companies to introduce new AI applications.
However, the Commission appears to recognize the risk of creating excessively prescriptive rules for AI that could place significant administrative burdens on the private sector. It should only consider new regulations in high-risk scenarios where there is clear evidence of consumer harm.
The EU already has a wide body of existing legislation that provides robust consumer protections and oversight of digital systems, including the GDPR, which already covers AI systems. The Commission should avoid requiring AI systems to be trained on datasets that conform to specific EU rules on traceability and data quality, as it would limit the available data for EU businesses and increase costs for consumers and companies.
In conclusion, the EU AI white paper presents an opportunity for the Commission to strike a balance between regulation and innovation. By taking into account the specificity and diversity of AI systems, encouraging industry best practices, and only imposing new regulations in high-risk scenarios, the Commission can support the development and uptake of AI while protecting consumers and ensuring a competitive market.
