What EU AI policy developments should businesses track in H2 2026? 

On 2 August 2026, AI Act obligations for GPAI models (General Purpose AI) entered into force - confirming the EU's position as the global frontrunner on AI and digital regulation more broadly. At the same time, two years on from Draghi’s report, Europe is still grappling with how to boost its standing in the global AI race, amidst growing concerns over its economic competitiveness and its "tech sovereignty" policy approach. This is giving rise to new initiatives that seek to drive investment and innovation while also favouring EU players over non-EU ones. These debates are still going on, with several key decisions expected in H2 2026. How they develop will play a key role in determining the regulatory environment for deploying and adopting AI in Europe for both European and international businesses. 

Source: Flint Global

The regulatory track: data, copyright and platform rules 

The EU AI Act remains the flagship regulatory instrument. Obligations on general-purpose AI model providers are already applying and being enforced by the European Commission's AI Office. They involve core obligations (technical documentation, downstream information, training summaries) and additional obligations for systemic-risk GPAI models (adversarial testing and systemic risk assessment, incident reporting, cybersecurity protections). Initially scheduled to apply this year, rules for providers of high-risk AI systems will not apply until 2027-2028 (2 December 2027 for systems listed in Annex III and 2 August 2028 for Annex I systems). While this timeline revision, adopted through the AI Omnibus, has given industry more runway, it has not reduced the volume of active legislative work. 

Two foundational questions are currently being resolved in parallel files. 

The first concerns data privacy. Even if the AI Omnibus has already been adopted, the Digital Omnibus — a separate file aimed at simplifying GDPR rules to support AI training — is currently in Parliament and Council discussions in Brussels. A central objective is to clarify the legal basis on which sensitive data can be used for AI training, a debate that has reignited the fears of privacy advocates about a retreat on data protection. This matters because the current ambiguity in GDPR interpretation has created significant uncertainty for AI labs seeking to train on European data at scale. Other similar initiatives aimed at streamlining existing regulations are being considered, such as the Digital Fitness Check, but it remains to be seen if it will tackle the question of applying more proportionality in regulations by assessing impact on competitiveness. 

The second concerns copyright. Over the AI Act debate, the EU has adopted transparency obligations on the use of copyrighted material for AI training under the AI Act. But the debate about rightsholders’ ability to assert their rights vis-à-vis GPAI model developers has remained, and the opt-out regime established under the 2019 Copyright Directive — which allows rightsholders to prevent their content from being used for text and data mining — is no longer considered sufficient. As a response, the European Commission is looking for potential solutions that would not involve reopening the Copyright Directive's text and data mining exception. Instead, Brussels is likely going to directly address the issue of licensing and remuneration, which has been at the core of the issue for the creative industry. In a recent call for evidence, the European Commission has started suggesting areas of evolution, that would include “facilitating licensing through mediation or arbitration or potential new measures to enable appropriate remuneration for creators, and strengthening protection of performers against AI-generated impersonifications". The debate is being revived in a more charged political context, with rightsholders mounting growing pressure for formal recognition of their contribution to AI training. How this is resolved will have significant implications for AI developers that rely on web-scraped or licensed content.  

On platforms, the DSA is increasingly relevant to AI. Popular standalone AI platforms — including general-purpose chatbots with large user bases — are likely to be designated as Very Large Online Platforms (VLOPs), triggering obligations around content moderation, algorithmic transparency and risk assessments. Beyond existing DSA rules, the European Commission is also considering additional measures to restrict social media access by age, which would ultimately extend to chatbots falling within the DSA's scope. 

Finally, one AI Act milestone has already landed. Article 50 — the transparency obligation for AI-generated content — started applying on 2 August 2026. It mandates machine-readable watermarks on AI-generated images, audio and video, and requires chatbots to self-identify at the moment of first contact. Ahead of this enforcement date, major AI providers signed the voluntary Code of Practice on Transparency of AI-Generated Content. Signatories that apply the Code’s provisions benefit from a presumption of compliance, though the underlying Article 50 obligations are mandatory for all providers regardless of whether they have signed. 

The competitiveness and sovereignty track: a push to boost AI adoption and its development  

Alongside the regulatory agenda, the European Commission has launched a series of initiatives designed to promote AI development in Europe. The political aim, as set out in the Apply AI Strategy presented by President Von der Leyen in October 2025, is to unlock the potential of generative AI in Europe by improving access to capital, compute capacity and data for European players. 

The centrepiece legislative file is CADA — the EU Cloud and AI Development Act. Its stated objectives are to triple EU data centre capacity, establish a four-level cloud and AI sovereignty framework, and reduce dependence on non-European technology providers. CADA is not only about infrastructure: it is also designed to create the right environment to train AI models in Europe, with specific provisions to unlock access to European data. 

Public procurement is another lever. Both CADA and the Public Procurement Act include provisions designed to favour EU-based and open-source AI solutions in public sector procurement — a significant channel given the scale of government technology spend across EU Member States.  

On compute, the Commission has published a call for proposals, with a €10bn envelope, and a target to reach €30bn with private capital added, to establish AI Gigafactory consortia across Europe, with a deadline for applications in November. These large-scale AI training infrastructure hubs are intended to give European developers access to the kind of compute capacity currently deployed at intensive pace in the US.  

Even with the EU Scaleup Europe fund recently set up, Europe remains far behind the hundreds of billions currently invested in US data centres and current EU initiatives must also be seen alongside Member States pursuing their own tech sovereignty agendas, such as France which has recently secured a €75bn investment by Softbank. To put this in perspective, the EU's total AI infrastructure commitment of €30bn — even with private capital — compares modestly with the €550bn Recovery and Resilience Facility deployed in 2021 to address the Covid economic shock, and that was spread across the entire EU economy. 

There is a palpable sense of urgency driving this work. The scale of AI investment being deployed in the United States has sharpened EU political attention on the risk of falling further behind. The Draghi report on EU competitiveness and the Letta report on the Single Market both identified this gap and called for accelerated investment and structural reform.  The current legislative programme represents the EU's attempt to operationalise those recommendations with sovereignty as a pillar of this strategy (see our recent blog post for further analysis: Europe's tech sovereignty push: What should businesses expect - CADA's public procurement provisions — a lever to promote EU players and boost competitiveness — seem modest in their ambition relative to the scale called for by the Draghi and Letta reports. Rather than deepening the single market and attracting global capital, the sovereignty agenda risks prioritising decoupling over competitiveness. Upcoming files could provide targeted responses, such as the one focused on enabling an EU Capital market (The Savings and Investment Union) and the discussion about the next MFF where the EU has a unique window to meet with Draghi’s recommendation of investing up to €750 billion to €800 billion per year. To succeed, Member States will have to politically align on how to equip the EU with the right innovation and industrial policy mandate and engage on the kind of ambitious structural reforms that have defined Europe's greatest achievements from the Schengen Area to the single currency. 

For businesses, the key takeaway is that the EU's AI policy agenda is broadening. Regulatory obligations are being streamlined across data, content and platforms, while industrial policy is creating new procurement opportunities and infrastructure commitments. The recommendations of the Draghi report have yet to be fully reflected in upcoming legislative files, and these discussions will need to take place both in Brussels and Member States. Monitoring both tracks simultaneously, and engaging early on the files still in motion, will be essential through the remainder of 2026. 

How can Flint Global assist? 

Flint Global advises companies, investors and financial institutions navigating complex regulatory, political and competitive developments across the UK, Europe, Asia-Pacific and Australia. Flint’s work spans tech regulation, innovation policies, digital markets, competition policy and geopolitical risk, helping clients anticipate policy change, assess strategic implications and position effectively in rapidly evolving regulatory environments. 

This blog was written by Sébastien Louradour, Senior Adviser at Flint Global. For further analysis on EU digital and AI policy, [contact us / see our related insights]. 

Enjoyed this blog? Sign up for our EU Monthly newsletter for monthly analysis of the most important policy and political developments across the European Union, providing timely insight into the issues shaping business and government.