Photo by Waldemar Brandt on Unsplash.
FORSEE’s mid-term policy brief identifies three principles running through eight tentative areas for European AI policy: stronger strategic self-determination, meaningful multistakeholder participation, and the integration of societal goals into AI standards, funding and procurement.
Recent AI safety incidents have highlighted growing risks. In July, hundreds of OpenAI’s AI agents launched a coordinated cybersecurity attack on the developer platform Hugging Face after breaching their development environment. Weeks later, during a similar test by Britain’s AI Security Institute (AISI), agents took on fake identities and contacted real people to try to persuade them to run malicious code.
These incidents illustrate one dimension of a broader challenge: as AI capabilities develop, societies need clearer criteria for what successful AI actually means. This is the question FORSEE has approached by examining how European institutions, SMEs, civil society, and public discourse define AI success.
Europe needs a broader definition of AI success
In order for AI to benefit the European economy and society, it needs to be trustworthy and societally sustainable. More than in the USA or China, leading European companies are concentrated in safety-critical industries, where AI adoption rests on guarantees about the reliability of the systems. European SMEs interviewed by FORSEE generally welcomed EU regulation as a way to counter Big Tech dominance, while stressing the need for clearer rules and practical compliance support.
The EU already possesses key regulatory instruments and supervisory frameworks for responsible AI development. In order to stay ahead, however, we must do more. FORSEE’s research finds both convergence and significant disagreement across European stakeholders.
Stakeholders broadly support binding regulation, worry about the risks of generative AI, and share an interest in protecting employment. At the same time, stakeholders differ over what AI success should mean, how digital sovereignty should be pursued and how bias should be addressed, while sustainability remains comparatively overlooked. After researching what AI success means, FORSEE has distilled its findings into tentative recommendations for European AI policy.
1. Strengthening the EU’s strategic self-determination in AI is a prerequisite for a strong regulatory framework
A strong regulatory framework depends on the EU having sufficient strategic capacity to set, implement and enforce its own rules. This means clarifying and enforcing the AI Act, giving practical support to SMEs, and reducing reliance on non-European infrastructure.
- Clarify the existing regulatory framework and support SMEs with compliance.
- Clarify how rights under the EU Charter of Fundamental Rights will be equally protected in AI governance.
- Define the role of Market Surveillance Authorities (MSAs) in identifying and remedying breaches of fundamental rights by AI systems.
- Support SMEs and the European AI industry with compliance by developing a dynamic register of AI-related harms, SME-specific guidance materials, and compliance tools for the AI Act.
- Strengthen digital sovereignty in line with EU values.
- As an addition to existing instruments, establish a European AI compute voucher scheme that grants SMEs subsidised access to high-performance computing, GPUs, and EU-hosted cloud resources.
- Improve SME access to high-quality and trusted data, including through the development of sectoral European Data Spaces.
2. Meaningful multistakeholder participation must extend beyond industry
Civil society organisations and other non-industry actors often lack the resources to participate effectively in AI governance. For AI to be broadly successful, these voices are exactly the ones that should be heard.
- Strengthen multi-stakeholder engagement in AI governance frameworks.
- Adopt binding rules on balanced representation within advisory and standard-setting bodies.
- Reinforce collective complaint and representation mechanisms, particularly in response to systemic risks such as labour precarity and surveillance.
- Enable capacity for civil society organisations (CSOs).
- Establish dedicated funding streams for CSOs, enabling them to develop technical AI expertise, and to advocate for marginalised groups’ rights.
3. Societal goals need to shape AI funding, standards and procurement
Societal goals, such as socially and environmentally sustainable development and bias mitigation, are currently treated as secondary intentions rather than binding conditions. How these goals are incorporated into governance matters for whether AI development aligns with broader European social and environmental objectives.
- Highlight, centre and address AI’s sustainability impact.
- Ensure that social and environmental considerations are integrated into binding standards. Integrate sustainability as a formal eligibility and evaluation criterion in AI funding programmes and public procurement frameworks.
- Support the development of EU-wide metrics and standards to assess the “hidden costs” of AI systems, and integrate these metrics into AI standards.
- Expand gender bias and discrimination mitigation mechanisms.
- Expand existing regulatory architectures and Fundamental Rights Impact Assessments (FRIAs) to explicitly address gender bias.
- Clarify the interaction between the AI Act, the GDPR, and existing anti-discrimination regulation at national and supranational levels.
- Create incentives for SMEs to include underrepresented groups in AI development.
These are preliminary directions: FORSEE will test and further substantiate them during the remaining stages of the project. Read more from the policy brief.
