Human Centric Generative AI made in Europe

Background and scope:

The rise of generative AI is astonishing. By seamlessly enhancing human abilities with machine capabilities, this next wave of AI may boost productivity in many sectors, create a new industry and also lead to profound socio-economic changes.

Furthermore, Generative AI is likely to revolutionise human-computer interaction, fostering more intuitive, conversational, and adaptive experiences.

Nevertheless, these advantages are not without some limitations. Current generative AI models function based on predictions rather than understanding, and their extensive capabilities and inherent risks are yet to be fully discovered.

The aim of this Challenge is to foster a European, human-centric approach to AI, tackling prevalent issues like transparency deficit and trust inadequacy. European AI start-ups have the potential to develop the next generation of generative AI models that embody EU values and guarantee Europe’s sovereignty in this critical field.

Specific objectives:

This Challenge aims to support the development of:

• foundation language and multimodal ‘frontier’ models that reach performances at least equivalent to the most powerful state of the art large generative models, capable of meeting the needs of European user industry, scientists, public sector and citizens;
• smaller foundation models with highly promising performance competing with frontier models in specific domains.

It is expected that the developed models go beyond the current state of the art in a way suitable for overcoming the current difficulties and limitations of this kind of tools. Examples of areas in which there could be relevant technological improvements include:

• Reliable content: Generative AI models minimising fictional elements;
• Transparency and traceability: Generative AI models allowing to trace the origin of the information provided.

The targeted applicants are primarily SMEs developing models themselves, but could also include SMEs providing innovative infrastructure, development tools, and critical support to the developers of generative AI solutions, in helping the efficient use of existing models while addressing specific issues such as hallucination or limited models knowledge.

The applicant must demonstrate a genuine commitment to developing and deploying “European-Value driven” AI. This European perspective should become a differentiating factor that will bring a competitive advantage to these companies, and also an important element to de-risk future investments.

Expected outcomes and impacts:

This Challenge is expected to reinforce the development of foundation models, which are “European-Value driven”, in line with the trustworthy and ethical principles as well as the (draft) AI Act.

The AI models developed and deployed under this Challenge are expected to comply with the EU concept for Trustworthy AI[[See Ethics guidelines for trustworthy AI (https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai)]] and the relevant ethical principles[[See Ethics By Design and Ethics of Use Approaches for Artificial Intelligence (https://ec.europa.eu/info/funding-tenders/opportunities/docs/2021-2027/horizon/guidance/ethics-by-design-and-ethics-of-use-approaches-for-artificial-intelligence_he_en.pdf )]] as well as the (draft) AI Act. In that respect, besides performances, due attention should be paid to data quality, transparency, privacy, and security.

In the mid and long term, it is expected to reduce dependencies and support European companies in leveraging the advances in generative AI enhance their products and develop new ones.

The selected beneficiaries will receive favourable access to European supercomputing resources for the training of their large foundation models in line with the access terms and conditions of the EuroHPC regulation. In addition, they may benefit from additional actions aimed at creating strategic partnerships with major industries or attracting further capital. In addition, opportunities may be explored to provide the selected beneficiaries with access to scientific datasets through the European Open Science Cloud or to provide users of the European Open Science Cloud with access to the tools developed by the beneficiaries.

Specific conditions

Any technology under this Challenge must be developed in a robust manner, paying specific attention to safety, security and ethics considerations in future applications.

In order to achieve the expected outcomes, and safeguard the Union’s strategic assets, interests, autonomy, and security, it is important to avoid a situation of technological dependency on a non-EU source, in a global context that requires the EU to take action to build on its strengths, and to carefully assess and address any strategic weaknesses, vulnerabilities and high-risk dependencies which put at risk the attainment of its ambitions. For this reason, beneficiaries of the grant component of Accelerator funding[['i.e. the legal entity which signs the grant agreement.]] must not be directly or indirectly controlled by a non-associated third country or a legal entity established in a non-associated third country other than such third countries or legal entities established in OECD member countries, Mercosur member countries countries with which the EU cooperates under a Trade and Technology Council, and countries with which the EU has a Digital Partnership [[OECD member countries have demonstrated a commitment to trustworthy and human-centric AI, as outlined by their adoption of OECD principles. Collaboration with India, Singapore, and Mercosur countries fosters not only technological advancements but also the ethical and secure deployment of AI aligned with European values. The Trade and Technology Council with India, the Digital Partnership with Singapore, and the Framework Cooperation Agreement with Mercosur, serve to mitigate the strategic and technological risks for collaboration with these countries.]].

Furthermore, in case of an investment support, specific safeguards may be introduced in the investment agreement (see Introduction, section on Economic Security).

Indicative budget: EUR 50 million.