EuroStack supports the Cloud and AI Development Act. But the first Council compromise would weaken key protections – here are our 19 requests to keep the Act strong

cloud and ai act eurostack

Position paper · 10 November 2026

EuroStack supports the Cloud and AI Development Act. But the first Council compromise would weaken key protections – here are our 19 requests to keep the Act strong.

Submitted by: EuroStack Initiative Foundation e.V., the industry-led initiative for European digital sovereignty

To: European Commission, Directorate-General for Communications Networks, Content and Technology

Re: COM(2026) 502 final, 2026/0138 (COD), Cloud and AI Development Act: feedback on the proposal

Executive summary

EuroStack supports the Cloud and AI Development Act[1]. For the first time, Union law would tell public bodies what degree of sovereignty to require when they buy cloud services, and would let them reward suppliers who build capacity in Europe.

This change of direction is under pressure: the first compromise text circulated in the Council weakens several of the protections the Commission proposed[2]. Procurement alone will not build a European digital industry; industry and private capital will[3]. They invest when demand is predictable[4], which the Act can provide by making clear that qualifying European supply will be bought.

Three principles run through our requests:

  • Control matters as much as location. A public body that cannot leave its supplier pays for it every year, whether or not any foreign law is ever invoked. The United Kingdom’s competition authority found that fewer than one cloud customer in a hundred changes provider in a year[5].
  • A requirement protects only if someone independent can check it. Otherwise an offer can pass as sovereign on paper while control stays elsewhere. That is sovereignty washing, for which verification is the remedy.
  • Buying European and interoperability go together. Open interfaces and portable data let a public body move its systems piece by piece, each move making the next one cheaper[6].
Our main requests

  1. Keep six protections of the proposal, four of which the Council compromise text would weaken (requests 1 to 6).
  2. Require level 3 or 4 for sensitive public activities (Article 30(3), request 7).
  3. Cost the “disproportionate cost” exception over the contract’s life, exit included (request 8).
  4. Make audits independent of the company audited (request 9).
  5. Let European contribution decide an award, on condition of three control tests (request 13).
  6. Make open source the default, with written justification for exceptions (request 14).
  7. Bind the Commission’s central purchasing to the sovereignty objectives (request 16).

Keep the proposal’s protections

The six provisions below should stay as the Commission proposed them. The press reports that the first Council compromise weakens the first four[2]. The last two are the proposal’s firm figures: a deadline and a target.

1Telemetry and metadata stay in the UnionAnnex II, point (c) at each level

Telemetry and metadata are the working data of a cloud service. The people who monitor, secure and repair a service work from them, so wherever they are processed is where much of the service is, in practice, run. The proposal keeps customer data, “including metadata and telemetry data”, in the Union at levels 1 to 3 “unless the public sector body explicitly requires otherwise”. At level 4, data identified as sensitive stays in the Union without exception.

The changes reported in the Council would turn the buyer’s explicit requirement into a mere permission, and would give level 4 an opt-out[2]. A public body that cannot easily leave its provider is in no position to refuse an export proposed in the provider’s standard terms, so a permission soon becomes the default. An exception for processing abroad on security grounds would raise the same difficulty in another form, since the reason is easy to give and hard for an auditor to test.

We ask

That point (c) be kept as proposed at every level. If a security exception is accepted at the lower levels, it should list the categories of data concerned, be checked in the audit, and stop at level 2. Level 4 should have no opt-out.

2The sanctions criterionAnnex II 2.1(g)(iv) and 3.1(g)(iv); Article 18(1)(c)

Sanctions are the most direct form of foreign leverage over a service, because they bind a provider whatever its customers want. At level 2, a provider controlled from a third country must therefore show that this control “is not exercised in a manner that obliges the audited provider to implement, enforce, give effect to, or comply with restrictive measures such as sanction regimes, embargoes, or any equivalent legal or administrative measures adopted by a third country”, unless those measures are legitimate under national or Union law.

Level 3 excludes such providers, except those of a third country that the Commission has associated under Article 18, to which the same criterion then applies. Article 18(1)(c) sets two conditions for that association. The third country must have no measures “to compel the cloud computing service provider to degrade or disrupt service continuity or provision”, and none that oblige the provider to apply sanction regimes or embargoes. The press reports that the Council text deletes the second condition[2].

The strength of the sanctions wording is that it can be checked in advance. Whether a country’s law can oblige a provider to apply that country’s sanctions is a matter of its statute book, settled once for every provider it controls. The condition on degrading or disrupting a service, left alone, could only be applied after the event, by arguing about what happened.

The sanctions against the prosecutor of the International Criminal Court in 2025 show how that argument goes. Court staff reported that Microsoft cancelled his email address[7]. Microsoft stated that the company’s contact with the Court “ultimately led to the disconnection of a sanctioned official from Microsoft services”, and also that it “at no point stopped or suspended its services to the ICC organisation”[8]. On Microsoft’s account no service to the Court was disrupted, so the remaining condition would not have caught the case. With an adequacy decision in place for the United States since 2023[9], Article 18(1)(a) is already met, which leaves the sanctions condition as one of the few that decide association.

We ask

That the sanctions wording be kept in each of these provisions. One correction is needed alongside. Annex II 3.1(g) lets a third-country-controlled provider be audited for level 3 “where the Commission has adopted an implementing act under Article 19”, but Article 19 concerns self-assessment and confers no implementing power, so the route cannot operate as drafted. We ask that the derogation be deleted or, if it is kept, that it refer to Article 18 with the conditions of Article 18(1) unchanged.

3The provider’s guarantee on vulnerability-reporting lawsAnnex II 1.1(g), 2.1(i)(iii) and 3.1(i)(iii)

This criterion is easy to mistake for a rule on access to data. Its subject is different and just as serious: whether a foreign state learns of flaws in services the Union depends on before they are fixed. A provider controlled from a third country “guarantees that there are no existing laws and practices in that third country, demonstrated by independent sources, that require the cloud computing service provider to report information on software vulnerabilities to authorities of that third country prior to those vulnerabilities being known to have been exploited”. Such laws exist. Since September 2021, China has required providers of network products to report vulnerabilities to the Ministry of Industry and Information Technology within two days[10].

The press reports that the Council text replaces “guarantees” with “demonstrates”[2]. The word matters because the criterion covers practices as well as laws. Laws are mostly public, so their existence can be shown either way. A practice can be secret, while no public source can prove that a secret practice does not exist. The provider knows whether one exists; the auditor does not. A guarantee therefore makes the party that knows answerable if such a practice comes to light, which a demonstration from the public record does not[11]. Without it, the criterion would be met by every provider with obligations that are not public, including the providers it was written to exclude.

We ask

That “guarantees” be kept.

4Defence, national security and the Commission’s role in setting levelsArticles 29 and 30

Member States decide, through risk assessments, which of levels 2, 3 or 4 applies to their sensitive public activities (Article 29(1)). The Commission’s role is what keeps those choices consistent across the Union. Its methodology must “specify how Member States use the highest level of assurance for the most critical public sectors activities including, but not limited to, defence” (Article 29(3)). The Commission may also set the level itself where a Member State’s choice “does not adequately address the public order concerns” (Article 29(5)). The press reports that the Council text removes defence data from the scope[2].

These provisions are what lift the most sensitive activities above the minimum. Without them, the minimum of Article 30(3) becomes the rule for most of those activities, and that minimum admits a provider controlled from a third country (request 7).

We ask

That defence and national security stay in the scope of Articles 29 and 30, and that Article 29(3) and (5) be kept as proposed.

5A firm migration deadlineArticle 29(6)

Where a risk assessment requires a public body to change service, it must migrate within a period that “shall not exceed 12 months”. The limit matters because a deadline that can be extended turns into a negotiation, in which the incumbent supplier holds the stronger hand: until the move is complete, the buyer cannot run its service without it.

We ask

That the 12-month limit be kept. If the co-legislators allow an extension for large systems, it should be granted once, for a stated period, with the decision and the reasons for it published.

6The 25% SME objectiveArticle 33(4)

Member States must pursue the objective that “at least 25% of their procurement for cloud computing services and AI systems be awarded to innovative SMEs”. A figure can be reported against, year by year and Member State by Member State. A formula such as “an adequate share” cannot be reported against, since no result would ever fall short of it.

We ask

That the figure be kept, with each national strategy under Article 7 reporting against it, using the series of request 19.

Reliable assurance levels

An assurance level is a promise to a public buyer about who can read, change or stop a service. The promise is worth as much as the level demanded for sensitive work, the exceptions allowed around it and the checks behind each claim.

7Level 3 or 4 for sensitive public activitiesArticle 30(3)

Article 30(3) requires “Union assurance level 2, 3 or 4” for public activities that contribute to public order: the NIS2 sectors, national security, internal security, border management, defence, justice and law enforcement. Levels 2 and 3 differ on one decisive point. At level 2, a provider controlled from a third country qualifies by documenting legal, technical and organisational mitigations (Annex II, 2.1(g)). At level 3, that control disqualifies it (3.1(g)).

Documented mitigations are cheap to write and very hard to audit. A requirement that can be met on paper and cannot be checked does not tell apart the provider a public body wants from the one it fears[12]. Applied criterion by criterion, the same test leaves three fifths of the Commission’s own sovereignty scoring framework unable to tell a foreign-controlled supplier from a European one[13]. The “2” in Article 30(3) is the single most consequential character in the Regulation.

We ask

That Article 30(3) read “level 3 or 4”, with the scope of sensitive activities drawn broadly.

8The “disproportionate cost” exceptionArticle 30(4)(c)

A contracting authority may set the whole framework aside where applying it “would require the contracting authority to procure services at disproportionate cost”. The term is not defined and no comparison is set. The exception must be “duly justified”, but the text does not say to whom, while nothing is published.

Read as most buyers will read it, the exception compares this year’s prices, and on that basis it will find the European offer disproportionate almost every time. It leaves out the cost of leaving the incumbent later: migrating data, rewriting integrations, retraining staff. For a five-year buyer of a leading office suite, the lowest estimates of that cost range from 0.4 to 0.9 years of spending[14]. A comparison without it would be arithmetically correct on the wrong quantity.

We ask

That Article 30(4)(c) require the comparison to cover the life of the contract, the cost of leaving included, and that each use of the exception be published and time-limited.

9Independent auditsArticles 19 and 20

Who chooses and pays the auditor decides how much an audit is worth. At level 1, the provider assesses its own conformity and issues an EU statement of conformity, by which it “shall assume responsibility for the compliance” of the service (Article 19(2)). At levels 2 to 4, the provider undergoes audits “at their own expense” by an auditing organisation it engages (Article 20(1)). Article 20(4) excludes the plain conflicts of interest: no other services to the provider in the twelve months around the audit, no audit of the same provider in the ten preceding years, no fee that depends on the result.

Those safeguards deal with an auditor’s ties to one client. They leave untouched the market the auditor sells in. An auditor chosen and paid by the company it examines competes for that company’s next contract, while the company remains free to take the business to a more accommodating firm. The credit rating agencies paid by the issuers they rated showed where that leads[15].

We ask

That, for levels 2 to 4, the national competent authority that evaluates the application under Article 17 assign the auditor from an accredited list, with the provider paying a fee the authority sets. The ten-year exclusion of Article 20(4)(a)(ii) should stay. The level 1 statement would be checked against public records: licences, code repositories and company registers. So would every declaration made for the award criteria of Article 32.

10Who controls a component over timeAnnex II 4.1(i)(ii)

A service is only as independent as the least independent thing it depends on. Each assurance level examines one provider at a time: the place of establishment, the location of infrastructure and data, the staff, the controlling owner. A service can meet every one of those criteria while depending on a component, such as an identity service or a piece of core software, that someone outside the Union can change or withdraw.

Only level 4 looks at this. Its criterion 4.1(i)(ii) requires the provider to demonstrate that a third country or a legal entity established in a third country “does not hold or exercise effective control over the design, development, maintenance, and evolution of those components or products”.

We ask

That this test apply from level 3. The software bill of materials already required at 2.1(i)(i) would be assessed along the chain of dependencies. Where a component’s maintenance is controlled from outside the Union, the migration plan the Annex already requires would name the alternative and what it would cost to use it.

11Open-source components treated no worse than proprietary onesAnnex II, criterion (j)

At levels 2, 3 and 4, criterion (j) requires controls, implemented and documented, on any component released under an open-source licence, whatever the origin. The parallel duty on proprietary components, in criterion (i)(ii), applies only when they are owned from a third country. The burden thus falls on the one kind of component whose source code anyone can inspect, and on the European companies that build services from it, while the dependency the Annex is concerned with goes untouched.

We ask

That the duty apply equally to open-source and proprietary components, according to who controls them.

12No lowering of the criteria by delegated actArticle 16(2)

Article 16(2) lets the Commission amend the assurance levels in Annex II and the evidence in Annex III by delegated act. The criteria required of a supplier to the public sector are an essential element of the Regulation. The EU cloud certification scheme shows how such criteria can be lost without any legislative decision. Requirements on ownership and on immunity from foreign law, present in its drafts since 2022, were dropped from the draft of March 2024[16].

We ask

For a non-regression clause stating that the power “is exercised for the purpose of clarifying or rectifying criteria, never of objectively lowering them”. Whatever criteria the final text closes to amendment, the others could be raised and never lowered.

Public purchasing

The sovereignty criteria decide who may sell to the public sector. The procurement provisions decide what public bodies actually buy, which is where demand for European supply is won or lost.

13European contribution that can decide an awardArticle 32

Public buying is the strongest tool the Union has to grow a European digital industry; Article 32 is where the Act takes it up. Contracting authorities must score “the tenderer’s contribution to the development of a European cloud and AI ecosystem”. Article 32(2)(d) then requires that score to be “ancillary and not decisive in the award of the contract”, while the recital suggests “a maximum weighting of 15 out of 120 points”. A modest price advantage makes up 15 points in 120; the text in any case forbids the score from deciding the award. France has proposed a minimum weight of 20% of the total score[6].

Three further points decide whether the preference rewards European control or only a European presence:

  • Scope. Article 32(1) applies to “innovative cloud computing services and AI systems”, a category the text does not define. Any buyer can place a purchase outside it.
  • Control. A supplier can employ thousands of people in the Union and still decide from elsewhere next year’s price, which features stay, when support ends and which law reaches the data. A European subsidiary reselling its parent’s technology passes tests on staff and head office, while the parent keeps those decisions[6].
  • Group level. Measured on a global group, a minimum share of staff in the Union excludes most large non-European vendors. Measured on a local subsidiary, it excludes almost none.

Scores on a list of criteria can offset one another, so a supplier shows its strong points and hopes the buyer overlooks the weak one. For a given service the weak point decides: if the maintenance team, the source code or the ability to leave sits outside the Union, staff numbers and a head office do not make up for it. EuroStack’s Buy European framework therefore separates the conditions a supplier must meet from the qualities a buyer scores[17].

We ask

That Article 32(2)(d) be deleted and the recital’s cap replaced by a minimum weight. Article 32 would apply to all cloud and AI procurement covered by the Regulation, without the word “innovative”. The contribution would be measured at group level.

A tenderer would earn the score only if it met three control tests: maintenance in the Union with access to the source code, core research and development in the Union, and a priced and tested exit. The remaining criteria would be scored. Among them, the score would credit core software released under an open-source licence, code open to audit, and contribution to the open-source projects the service depends on. That credit would follow from the licence and the contribution, wherever a project is hosted and whoever else contributes to it.

A preference reaches only new contracts, while most public spending renews contracts the buyer cannot easily leave. Requests 5, 8 and 14 address that larger share.

14Open source as the defaultArticle 41

Article 41 is titled “Promoting open source solutions and open source first”, but its operative sentence asks the Union and the Member States only to “take the necessary measures to encourage” public bodies to use open standards and open-source components “when building their cloud and AI ecosystem or stack, taking into account functionalities, including security, total cost, and other relevant, duly justified objective criteria”. The duty is to encourage, it covers only the cloud and AI stack, and its last clause lets an authority justify almost any choice after the event[3].

France has run this experiment for ten years. Article 16 of its 2016 law for a digital republic tells public administrations to encourage free software and open formats, with the same verb. A decade on, the State’s largest software contracts still go to the same few vendors[18]. An open-source licence gives a public body the freedom to hire someone else to maintain the software if the supplier fails or changes course, which keeps the cost of leaving low[14]. A duty to encourage, which no one checks, does not secure that freedom.

We ask

That Article 41 read “require, unless justified in writing”. The justification would state why no suitable open-source solution fits, compare costs over the life of the contract with the cost of leaving included, be read by a named authority before the decision, and be published. The duty would bind the body that signs the contract. It would apply whether a capability is bought as a licence or as a service, so that the same dependency cannot return as a subscription.

15Software paid for with public moneyArticle 42

Article 42 sets the catalogue through which a public body shares software once it decides to share it. It obliges no body to share software developed with public money, and covers only software to which the body itself holds the rights. Software developed under contracts in which the contractor keeps the rights, a common case, stays outside it[3].

We ask

That Article 42 make software developed with public money available under an open-source licence unless a stated ground prevents it, and that procurement contracts leave the public body the rights needed to do so.

16The Commission’s central purchasingArticle 38(5)

This is the largest single lever in the Regulation, and the least specified. Under Articles 37 to 39, the Commission may buy data centre services, cloud services, software and AI systems for Union entities and Member States, and a body buying this way is deemed to have met its own procurement obligations (Article 39(1)). Whatever the Commission’s framework contracts require becomes the de facto European standard. Whatever they omit is omitted everywhere at once, without any acceding authority having to decide anything. The Commission has already shown the way by applying the Cloud Sovereignty Framework to one of its own procurements, the Cloud III tender[19].

We ask

That Article 38(5) require the strategic direction of each procurement to show consistency with Article 1(1)(d) and with the assurance levels of Article 30, and that this assessment be published.

We also ask for a change-of-control clause in the framework contracts. An assurance level records who controls a supplier on the day of the audit, and control can change the next day. The Abilian Econ Lab has documented 24 European technology companies that passed under foreign control through acquisition, buyout, a move abroad or a venture operating a foreign provider’s product[20]. A change in control of the supplier, as defined in Article 2(21), would trigger a new assessment and give the contracting authority a right to terminate. Any change in a holding of 5% or more of the capital or voting rights would be notified to it, the threshold audit criterion G of Annex III already uses.

Public support and monitoring

Public money for data centres and research should widen the choices open to European buyers. The Union should also be able to see whether the policy is working.

17Conditions on acceleration zones and strategic projectsArticles 11, 13 and 14

Data centre projects in acceleration zones and strategic projects receive faster permits and other support, with few conditions attached. Article 14(1) grants strategic status to a project meeting “at least two of the following” five criteria, so adding openness as a sixth would change nothing: a project could qualify without it. Switching and portability rights already exist in Chapter VI of the Data Act[21], so restating them would add no obligation either. The missing element is proof that those rights work in practice.

We ask

That a project receiving these benefits be required to show compliance with Chapter VI of the Data Act through a tested exit, for the life of the designation: data extracted, formats validated, a migration path actually exercised. Article 11(2), which already forbids discriminatory allocation of resources and foreclosure, is the natural place for it.

18Pillar 1 governanceArticle 6 and Annex I

Who writes the calls decides where research money goes. Article 6(1) entrusts the Cloud and AI Leadership Initiatives to “the Commission and the Member States and, where relevant, to joint undertakings or any other structures capable of achieving those objectives”. The Commission’s own presentation of Pillar 1 shows what would change[22].

Today, cloud and edge calls are “drafted internally by Commission services (DG); approved by the Programme Committee (Member States)”. Under the Act, the work programme would be “co-designed by the JU Governing board, where the Commission and industry/State members vote together”, and the “Business association shapes priorities, grand challenges and calls”. On such a board the private side, represented by industry associations, holds half the votes. The presentation says the board “includes SME representatives”, without reserving them any seat. The grand challenges that calls implement are set in Annex I, which the Commission may amend alone by delegated act (Article 6(4)).

We ask

For three changes. An openness and dependency-reduction requirement would be written into Annex I and excluded from the delegation. Any entrusted structure would publish how its work programme serves Article 1(1)(d) and the objectives of Article 4, and SMEs and open-source providers would hold reserved voting seats. Article 6(3), which names the Digital Europe Programme, would add “and successors”, since that programme is due to be folded into the European Competitiveness Fund from 2028[22].

19Measuring the resultArticle 15

Article 15 has the Commission monitor compute capacity, demand and the capacity gap. The accompanying open-source strategy adds fourteen indicators, such as the number of public authorities using Commission guidance and the number of companies supported[23]. All fourteen measure effort; none measures the result. The figure that would show whether the policy works is the share of public spending on cloud, software and AI that goes to suppliers recognised at each assurance level, by Member State and by year.

We ask

That Article 15, or the national strategies of Article 7, require that series, with a target path.

EuroStack Initiative Foundation e.V. · eurostack.eu

References

  1. European Commission, Proposal for a Regulation establishing a framework of measures for strengthening Europe’s cloud and AI ecosystem (Cloud and AI Development Act), COM(2026) 502 final, 3 June 2026, with Annexes I to III. eur-lex.europa.eu
  2. Hartmann, T. and Henning, M., “Ireland makes pro-US Big Tech amendments to EU sovereign cloud law”, Rapporteur, 2 November 2026 (subscription required). rapporteur.com
  3. EuroStack, “EC’s Tech Sovereignty Package – UPDATE”, 8 June 2026. eurostack.eu
  4. Caffarra, C., Charrier, P.-H., Sohns, A. and Stojanovski, O., EuroStack: Organizing “Demand Pull”, Not Just “Supply Push”, EuroStack, 2026. eurostack.eu
  5. Competition and Markets Authority (United Kingdom), Cloud services market investigation: summary of final decision, 31 July 2025, paragraph 24. gov.uk
  6. EuroStack, “Made in Europe, decided elsewhere? A EuroStack reading of the French paper on value creation in CADA”, 30 September 2026. eurostack.eu
  7. Associated Press, “Trump’s sanctions on ICC’s chief prosecutor have halted tribunal’s work, officials and lawyers say”, 15 May 2025. pbs.org
  8. van Trigt, M., “Microsoft: ‘Geen diensten aan ICC-organisatie gestopt'”, Binnenlands Bestuur, 20 May 2025; Microsoft’s statement quoted in our translation. binnenlandsbestuur.nl
  9. Commission Implementing Decision (EU) 2023/1795 of 10 July 2023 on the adequate level of protection of personal data under the EU-US Data Privacy Framework. data.europa.eu
  10. Ministry of Industry and Information Technology, Cyberspace Administration of China and Ministry of Public Security, Regulations on the Management of Network Product Security Vulnerabilities, July 2021, in force 1 September 2021. Summary in English: “Chinese government lays out new vulnerability disclosure rules”, The Record, July 2021. therecord.media
  11. Grossman, S. J. (1981). The informational role of warranties and private disclosure about product quality. Journal of Law and Economics, 24(3), 461–483.
  12. Fermigier, S. (2026a). Screening for Sovereignty: Optimal Procurement when Compliance is Cheap to Fake and Costly to Verify. Draft working paper v0.20, 2026-10-21. Abilian Econ Lab. econ.lab.abilian.com
  13. Fermigier, S. (2026i). Ask Me Something Hard: Why Three Fifths of Europe’s Sovereignty Score Cannot Tell Suppliers Apart. Draft position paper, 2026. Abilian Econ Lab. econ.lab.abilian.com
  14. Fermigier, S. (2026b). Exit, Pursued by a Bill: Lock-in Rents and Portability in European Cloud and Software. Draft working paper v0.14, 2026-09-23. Abilian Econ Lab. econ.lab.abilian.com
  15. Bolton, P., Freixas, X. and Shapiro, J. (2012). The credit ratings game. Journal of Finance, 67(1), 85–111. doi.org
  16. Foo Yun Chee, “EU drops sovereignty requirements in cybersecurity certification scheme, document shows”, Reuters, 3 April 2024.
  17. EuroStack, A Proposed Framework for a “Buy European” Regulation of Strategic Digital Procurement, 29 September 2025. eurostack.eu
  18. Fermigier, S. (2026j). Loi République numérique : dix ans d’« encouragements », pour quels résultats ? Maîtrise, indépendance et logiciel libre dans les systèmes d’information de l’État, de la loi de 2016 au règlement européen de 2026. Document de travail provisoire v0.2, 2026-09-22. Abilian Econ Lab. econ.lab.abilian.com
  19. European Commission, Cloud Sovereignty Framework: Implementation guidance, Cloud III Dynamic Purchasing System, with the annexed sovereignty assessment calculator. The framework itself is version 1.2.1, DG Digital Services, 20 October 2025. commission.europa.eu
  20. Fermigier, S. (2026f). Sovereignty as a Flow: Capture and the Durability of European Technology. Draft working paper v0.7, 2026-09-23. Abilian Econ Lab. econ.lab.abilian.com
  21. Regulation (EU) 2023/2854 (Data Act), Chapter VI on switching between data processing services. data.europa.eu
  22. European Commission, DG CNECT, Cloud and AI Development Act: Pillar 1 R&D&I, and status of next MFF negotiations. Presentation to the Cloud Alliance industrial working group, 2026. Not publicly available.
  23. European Commission, Communication on European Tech Sovereignty, accompanied by an EU Open Source Strategy, COM(2026) 503 final, 3 June 2026, Annex I. eur-lex.europa.eu

Share the Post:

Related Posts