Ada in Europe

Through our Brussels office, our work in Europe considers how the interests of people and society are met by data, digital and AI regulation. To do this, we provide evidence-based research, including technical research.

To find out more about our work, email our European Public Policy Lead Connor Dunlop.

The current environment

An overarching aim of the EU’s digital strategy is to rebalance power between big tech companies and society by addressing market imbalances and placing the responsibility and accountability on the large platforms known as ‘gatekeepers’.

EU digital legislation is key to this approach. The Digital Services Act sets rules on intermediaries’ obligations and accountability, while the Digital Markets Act seeks to ‘ensure these platforms behave in a fair way online’.

‘Gatekeepers’ are platforms that have a strong economic position, are active in multiple EU countries and act as a gateway between businesses and consumers in relation to core platform services. The European Commission has designated six companies as gatekeepers: Alphabet, Amazon, Apple, Meta, Microsoft and ByteDance.

Similarly, the EU’s data strategy centres on providing more equitable access to data through mechanisms such as data sharing and interoperability via the Data Act and Data Governance Act.

Other legislation will provide further protection. The AI Act will have a preventative scope, assigning obligations across the whole AI supply chain including post-market deployment. The AI Liability Directive will have a compensatory scope, that is, providing recompense if harm occurs.

Ada’s approach

The impact of initiatives – such as the Digital Services Act and the Digital Markets Act, and the AI Act and AI Liability Directive – will depend upon effective implementation and enforcement, including through the GDPR .

Our work currently focuses on regulatory capacity and institutional design. For example, we supported the EU institutions in conceptualising a strengthened ‘AI Office’ (rather than an ‘AI Board’ as originally proposed in the EU AI Act).

Our priorities for policy research

1. Governance including the EU AI Act and AI liability

As the EU AI Act has been adopted into law, we will study its implementation, including how its recitals and articles should be interpreted, to help ensure the Act meets its objective of protecting people and society from potential harms. Our previous work in this area includes convening, policy analysis and engagement on the AI Act. See our expert explainer, policy briefing and expert opinion.

The next big question for policymakers in the EU – and the UK – is how to allocate legal liability throughout an AI system’s supply chain. Ada has commissioned expert analysis around this, including  how liability law could support a legal framework for AI.

In 2023, to support policymakers, we also published a paper on risk assessment and mitigation in AI; and a paper on how governments can reduce information asymmetries through monitoring and foresight functions.

Our three-year programme on the governance of biometrics provides legal analysis and public deliberation on issues of safeguards, use and risks of biometric data and technologies.

2. Ethics and accountability in practice

We monitor, develop, pilot and evaluate mechanisms for public and private scrutiny and accountability of data and AI. This includes our research on technical methods for auditing algorithmic systems; a collaboration with AI Now and the Open Government Partnership which surveyed algorithmic accountability policy mechanisms in the public sector; and a pilot of algorithmic impact assessments in healthcare.

3. Rethinking data and rebalancing power

Rethinking data and rebalancing power reconsidered the narrative, uses and governance of data. The expert working group – co-chaired by Professor Diane Coyle (Bennett Institute of Public Policy, Cambridge) and Paul Nemitz (Director, Principal Adviser on Justice Policy,  EU Commission, and Member of the German Data Ethics Commission) – sought to answer a radical question:

What is a more ambitious vision for the future of data that extends the scope of what we think is possible?’

4. AI standards and civil society participation

Both the EU AI Act and forthcoming UK legislation will rely heavily on the creation of AI standards by international bodies like ISO.

Ada has previously researched how these standards processes can include more civil society voices as well as other innovations that might improve democratic control. We also hosted an expert roundtable on EU AI standards development and civil society participation.

5. Foundation model governance

We first called for the inclusion of foundation models (or ‘general-purpose AI models’) in regulation in early 2022. Since then, our work has increasingly focused on the unique risk profile of foundation models, and the challenges posed for policymakers.

Our work to support both governments and the wider ecosystem in understanding the AI value chain and the (potentially) critical role of foundation models began with establishing a shared terminology through our explainer. We also identified the key considerations for public-sector deployment of models like OpenAI’s GPT-4 and its application interface, ChatGPT.

We have also explored if foundation model governance can learn from the regulation of similarly complex, novel technologies in the life sciences sector. This includes regulatory powers and potential mechanisms such as pre-market approval for foundation models.

Image credit: DKosig