Speaking of EU legislation and.. of sausages making. Still today there is no way of knowing how EU “Safe Countries” legislation was made ..

By Emilio De Capitani

Setting the ..table

“Laws are like sausages. It is best not to see them being made” is the most famous Otto Bismark’s quote, (even though, there’s no evidence, that Bismarck ever said it, much less coined it…). This statement may well come to mind when looking at the procedure followed by the EU legislators when preparing, negotiating and adopting the legislation on “Safe Countries”. The legislative act is Regulation (EU) 2024/1348 establishing a common procedure for international protection in the Union. It consolidates the first list of EU “Safe countries” established by Regulation (EU) 2026/464 as well as the new rules defining the concept of ‘safe country’ eg Regulation (EU) 2026/463

It is worth noting that these texts creates some “legal presumptions” whose main justification is to make easier to repatriate or expel both irregular migrants and people who have not been granted the right to asylum[1] These legal presumption which amounts to consider as unfounded most of the request for protection of the people trying to enter the EU from these third countries and because of this it significantly restricts the rights of people who are extradited, removed or turned back. It will be up to them to prove in very short deadlines that, in exceptional circumstances, the concept of a safe country of origin should not apply in their specific case (see paragraph 5 of Regulation 2026/464)[2] . Furthermore, the right to judicial protection, guaranteed by Article 47 of the Charter of Fundamental Rights of the European Union, would be severely restricted, as judicial appeals must be lodged within very short time limits, would not automatically have suspensive effect and, in the event of a negative decision, the individual would not be permitted to remain legally within the territory of the EU.

On these aspects it is worth recalling the art.52.1 of the EU  Charter which states “Any limitation on the exercise of the rights and freedoms recognised by this Charter must be provided for by law (emphasis added) and respect the essence of those rights and freedoms.  Subject to the principle of proportionality, limitations may be made only if they are necessary and genuinely meet objectives of general interest recognised by the Union or the need to protect the rights and freedoms of others.limitations to fundamental rights may be adopted “

Has the principle of legislative transparency been respected in the Safe Countries case ?

In the following lines I will focus notably on some principles of the EU Legislation which in my opinion have been violated in this case.

It is self evident that these three Regulations have been adopted under the joint responsibility of the European Parliament and of the Council [3]following the so called “ordinary procedure” (art.294 TFEU) and are therefore EU “legislative acts” (art.289 TFEU). However it is worth recalling that since Lisbon art.15.2 TFEU has established the obligation of  legislative transparency which covers not only the votes (as it was the case before Lisbon) but also debates and negotiating preparatory documents. Unsurprisingly the CJEU has declared in 2018 that “transparency is inherent to legislative procedure” as legislative transparency is the main expression of EU democratic nature. As such, the co-legislator obligation of proactively publishing legislative preparatory documents shall not depend from the aleatory condition that citizens may submit a request for access to these documents. Proactive publication of legislative preparatory documents paves also the way for citizens’ timely participation to the EU decision making process as granted by art.10.3  11 of the TEU.

Building on the CJEU jurisprudence on Safe Countries ..

Furthermore in the case of the EU legislation defining the “safe countries” the general obligation of legislative transparency foreseen by art.15.2 TFEU has already been strengthened by the CJEU jurisprudence. According to the judgment of the Court of Justice of the European Union in Alace and Canpelli (in joined cases C-758/24, paragraphs 86–87, and C-759/24), “ the information justifying the designation of a third country as ‘safe’ – and on which a legal presumption may be based – must be precise and accessible not only to the public but, in particular, to the person seeking asylum or being refused entry; and, above all, it must be verifiable by any court called upon to rule on the matter[4](emphasis added). This jurisprudence was referring to national legislation dealing with “safe countries” and in my opinion is transferrable also to EU legislation. Moreover Article 61 of the same ‘Procedures’ Regulation, provides for the use of ‘relevant and available’ sources in order to verify whether a country can be defined as safe.  In other words, the designation of a country as ‘safe’ cannot be the result of mere political discretion, as it affects the exercise of a fundamental right (the right to political asylum) and because it may limit the prohibition on expulsion laid down in the Charter of Fundamental Rights of the European Union.

Unfortunately, it is by no means clear from the text and the preparatory work on the European legislation concerning safe countries what information the legislator relied on in designating several third Countries as ‘safe’. This is notably the case of Regulation (EU) 2024/0464, which, despite affecting the exercise of fundamental rights, did not undergo an impact assessment, as required by the 2016 ‘Better Law-Making’ Interinstitutional Agreement. Instead it merely refers to information gathered by the European Asylum Agency (EUAA) which may be considered the real impact assesment in this case. As such, following the 2018 “ClientHeart” jurisprudence on the accessibility of legislative impact assessment  it should had been, at least partially disclosed.  No way: when in 2024 I submitted to the EUAA a request of access to that information I received a negative answer for procedural reasons. Then I challenged the refusal of disclosure before the General Court (case T-661/25).[5] On my side the history could had be ended here awaiting for the General Court Ruling but something, unexpected, happened and may now re-open the issue of legislative transparency of that EU legislation on Safe Countries even before the CJEU.

India is a “safe Country” ? No according to the association “Sikhs for Justice”

It is worth noting that the EU has declared, that INDIA is a “safe” Country  for all its territory and for all the categories of people living there notwithstanding Article 63 of the ‘Procedures’ Regulation would had permitted to exclude parts of the territory or specific categories of persons. The latter approach has been adopted, for example, by the legislation of the Netherlands, which states: “The designation of this country as a safe country of origin does not apply to: • people originating from the ‘union territory’ of Jammu and Kashmir; • religious minorities who have experienced problems because of belonging to that religious minority; • Dalit women and girls; • journalists; • persons who have been critical of the Indian government and government policy, and have experienced problems as a result (for example, human rights activists, academics and demonstrators).” ([6] )

However these reservations, which, if mirrored at EU level would allowed for greater protection of certain categories of people and citizens of certain Indian states, can no longer be invoked because the new EU-level general declaration ([7]) declaring India as safe takes precedence over national declarations. However, a number of questions may arise, such as: which of the two declarations more accurately reflects the reality on the ground in India? On what facts was the Netherlands’ decision based, and on what other elements the European legislator has decided that certain safeguards intended for specific categories of people and specific regions of India can no longer be invoked?

This may explain why a representative of the Sikhs for Justice association has recently submitted a request of access to documents to the European Parliament and to the Council of the European Union.

Both the EP and the Council have, at least as a first reaction, refused to disclose the EUAA information.

How EU legislative transparency may be granted by invoking “classified” information?

On one side it has been evoked the risk for EU external relations even though what is at stake is the autonomy of EU “internal” legislation and not international negotiations.

On the other side both the institutions have declared that the Asylum Agency has ‘classified’ these information as ‘RESTREINT’ and that the EP and the Council have retained that classification.

This is quite disturbing because first the only institutions which may “classify” EU information according to art.9 of Regulation 1049/01 are at very large extent the Commission, the EP and the Council which were the only institutions covered by art.255 TCE on which Regulation 1049/01 was founded. Moreover it emerges very clearly from the legislative preparatory work of Regulation 1049/01 that art.9 was as an exceptional regime for information covering only EU and MS defence policies ([8]) and clearly not information dealing with a Country general condition of “safety” as in the this case.

It is at least unfortunate that the this EUAA ‘classification’ has been  subsequently endorsed by the Commission when the legislative proposal and, later, by the Parliament and the Council in their capacity as legislators. However even if endorsed by the EP and the Council this choice  appears open to criticism, at the very least, in terms of both form and content. It is open to criticism in terms of form, insofar as the ‘RESTREINT’ level is not provided for in Article 9 of Regulation (EC) No 1049/01, the only European legislation dealing with ‘sensitive information’ that may be shared only with authorised persons (‘need to know’). But it is open to criticism above all on the merits, as two scenarios could be envisaged:

  1.  either this information proves that India is a safe country, in which case it is difficult to see what negative consequences a possible disclosure might have,
  2. or, conversely, it proves the opposite, in which case its ‘classification’ would violate European law on two counts.

Indeed, by declaring a country to be safe, despite independent information to the contrary, the European legislator would be exercising its remit in an arbitrary manner (which should instead be based on independent and objective information, precisely because it affects the exercise of a fundamental right such as the right to asylum and compliance with the prohibition on refoulement, as laid down in the Charter of Fundamental Rights).

Above all, however, the ‘classification’ of the information used to justify the concept of a ‘safe country’ meant that its contents could not be debated publicly during the legislative negotiations, particularly in the European Parliament which denounced this situation in its resolution on access to documents.[9]  

Conclusion

Invoking classified information to justify the non disclosure of legislative preparatory documents seems to me a clear breach of the principle of legislative transparency laid down in Article 15(2) of the Treaty on the Functioning of the European Union, and risks, unfortunately, of bringing the European legislator closer to that of North Korea[10] .

The procedure for access to documents, initiated by the Sikhs for Justice Association, is still ongoing, so it is not yet clear whether the European Court will be called upon to settle the matter, but in the meantime, the Sikhs for Justice association has taken steps to limit the damage caused by the new European regulations, including through ‘participatory’ channels, by publishing an open letter and by submitting a petition to the European Parliament calling for the suspension and amendment of India’s designation as a ‘safe country’ (as foreseen by Article 63 of the ‘Procedures’ Regulation)

It now remains to be seen whether the legality of designating India as a ‘safe’ country will be reviewed through the courts or following a legislative review process. Whatever the future of the European legislation may be, it is certainly commendable and noteworthy that representatives of civil society, such as the organisation Sikhs for Justice, are fighting so resolutely not only for the respect of their own fundamental rights, but also for better quality legislation and European democracy, in accordance with the principles of democracy, the rule of law, the protection of minorities, solidarity and non-discrimination, upon which the European Union is founded.


[1] Article 41(1)(e) and Article 3(a) concerning accelerated procedures

[2] see: https://eur-lex.europa.eu/legal-content/IT/TXT/PDF/?uri=OJ:L_202600464

[3]  As required by see art 14 and 16 TEU.

[4] SEE: “(86 ) However, the effectiveness of judicial review regarding compliance with the substantive conditions set out in Annex I to Directive 2013/32 presupposes that the court before which the case is brought has access to the sources of information on the basis of which the competent national authority designated the third country in question as a safe country of origin. That requirement of effectiveness also implies that the court in question must be able to verify whether that designation complies with the substantive conditions set out in Annex I to that directive, taking into account any other information it may have gathered itself, whether from public sources or from sources which it has requested one of the parties to the proceedings before it to produce, provided that, on the one hand, the reliability of that information has been established and, on the other hand, in accordance with the principle of the right to be heard, those parties have the opportunity to submit their observations on that information. (87 ) It follows that Member States are required, under Article 46(3) of Directive 2013/32, to adapt their national law in such a way as to ensure sufficient and adequate access to the sources of information on which they have relied in designating safe countries of origin. Such access must enable an applicant for international protection originating from such a country, and the national court hearing an appeal against a decision concerning the application for international protection, to familiarise themselves effectively with those sources of information in accordance with the case-law referred to in paragraph 78 of this judgment.”

[5] Even though the main issue at stake in this case is the fact that I have submitted my confirmatory application with two days delay that EUAA has considered mandatory and I consider only of ordinatory nature.  

[6] See Council document WK 8158/2025, pp. 23–24

[7]  According to Regulation (EU) “With regard to India, according to information from the Asylum Agency, nine Member States currently designate it as a safe country of origin at national level, and in 2024 the recognition rate at Union level for applicants from that country was 2 per cent. The country has ratified the main international human rights instruments. India is a constitutional republic with a parliamentary democracy. There is no evidence of expulsion, removal or extradition of Indian nationals to countries where they would be at risk of being subjected to the death penalty, torture, persecution or other inhuman or degrading treatment. In general, there is no real risk of serious harm within the meaning of Article 15 of Regulation (EU) 2024/1347. Although India retains the death penalty in its criminal law and has not signed the Second Optional Protocol to the International Covenant on Civil and Political Rights, which aims to abolish the death penalty, in practice the death penalty has not been carried out since 2020. India has signed the Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment. There are no armed conflicts in India and, consequently, there are no threats arising from indiscriminate violence in situations of internal or international armed conflict. In general, there is no persecution in the country within the meaning of Article 9 of Regulation (EU) 2024/1347.”

[8] Following the adoption of art.9 the European parliament accepted to withdraw a Court Case dealing with an agreement between the Council and NATO (so called “Solana agreement”)

[9] “..calls, in particular, on the European Union Agency for Asylum (EUAA) to grant the European Parliament timely access to its documents, in particular information and analysis on safe countries of origin and safe non-EU countries..” Par 16 of  European Parliament resolution of 10 March 2026 on public access to documents – report covering the years 2022 – 2024 (2025/2137(INI)) (C/2026/4005)

[10] As stated in plenary by Graham Watson, Chair of the European Parliament’s Committee on Civil Liberties, on May 2nd,2000 when voting on the Access to documents last compromise.

AI liability rules: a blocked horizon?

By Michèle Dubrocard[1]

February 2025

Today, no one challenges the potential benefits offered by AI for individuals and society in general, but also the existence of serious risks, some of them already identified, others likely to emerge. Let’s have in mind the conclusions of the first International AI Safety Report[2] which, focusing on general-purpose AI, recognizes that ‘there is a wide range of possible outcomes even in the near future, including both very positive and very negative ones, as well as anything in between’. 

So, when the European Commission issued on 28 September 2022 its Proposal for a Directive on adapting non-contractual civil liability rules to artificial intelligence (AI Liability Directive- AILD), it raised a lot of hope among all those concerned about the potentially harmful consequences of the use of AI systems. These hopes were confirmed by the objective expressed in the explanatory memorandum of the Proposal, namely ‘ensuring victims of damage caused by AI obtain equivalent protection to victims of damage caused by products in general’[3].

More specifically, the Commission seemed determined to take into due consideration the imbalance between the providers and deployers on the one hand, and the affected persons on the other. Indeed, referring to Member States’ general fault-based liability rules, Recital 3 of the Proposal recognizes that ‘when AI is interposed between the act or omission of a person and the damage, the specific characteristics of certain AI systems, such as opacity, autonomous behaviour and complexity, may make it excessively difficult, if not impossible, for the injured person to meet this burden of proof’.

Alas, the rules proposed by the Commission did not meet the expectations raised by the announced objective (I). Even worse, the Commission seems to have definitively shelved its project (II), leaving the door open to what it itself had criticized: the co-existence within the EU of ‘27 different liability regimes, leading to different levels of protection and distorted competition among businesses from different Member States’[4].

I- A disappointing Proposal

The Proposal of the Commission did not challenge the choice of a fault-based regime, but instead mainly focused on two rules, aiming at alleviating the burden of proof, which remains on the victim. What are these two rules?

– The disclosure of evidence:

According to Article 3(1) of the Proposal, a court may order the disclosure of relevant evidence about specific high-risk AI systems that are suspected of having caused damage. However, the requests should be supported by ‘facts and evidence sufficient to establish the plausibility of the contemplated claim for damages’ and the requested evidence should be at the addressees’ disposal. Article 3(3) provides that the preservation of such evidence may also be ordered by the court.

However, the disclosure may be ordered by a court only to ‘that which is necessary and proportionate to support a potential claim or a claim for damages and the preservation to that which is necessary and proportionate to support such a claim for damages’. Article 3(4) specifies that ‘the legitimate interests of all parties’ must be considered by the court, when   determining whether an order for the disclosure or preservation of evidence is proportionate. Moreover, the person who has been ordered to disclose or to preserve the evidence must benefit appropriate procedural remedies in response to such orders.

Article 3(5) introduces a presumption of non-compliance with a duty of care: when, in a claim for damages, the defendant fails to comply with an order by a national court to disclose or to preserve evidence at its disposal, the national court shall presume the defendant’s non-compliance with a relevant duty of care. That presumption remains rebuttable.

– The presumption of causal link in the case of fault:

Article 4 of the Proposal provides, under certain conditions, a presumption of a causal link between the fault of the defendant and the output produced by the AI system or the failure of the AI system to produce an output, that gave rise to the relevant damage.

However, the claimant has to prove the fault of the defendant, consisting in the non-compliance with a duty of care laid down in Union or national law directly intended to protect against the damage that occurred. He/she also has to prove that the AI system gave rise to the damage. There is another condition, related to the likelihood, based on the circumstances of the case, of the fault’s influence on the output produced by the AI system or the failure of the AI system to produce an output.

Moreover, the presumption shall not be applied if the defendant demonstrates that sufficient evidence and expertise is reasonably accessible for the claimant to prove the causal link. At last, in the case of a claim for damages concerning an AI system that is not a high-risk AI system, the presumption shall only apply where the national court considers it excessively difficult for the claimant to prove the causal link. Here also, the presumption is rebuttable.

– The limitations of the rules:

It follows from these provisions that the impact of the two rules laid down in the Proposal is limited by numerous conditions. In particular, the new mechanism of disclosure of evidence would be limited only to high-risk AI systems. Similarly, the presumption of causal link would mainly apply to high-risk AI systems, except where, according to the national judges, it would be excessively difficult for the claimant to prove the causal link.

In any case, the Proposal is based on a fault-based liability regime, which means that victims would still have to prove the fault or negligence of the AI system provider, or deployer. As noted by the EDPS in its own-initiative opinion[5] of 11 October 2023, ‘meeting such a requirement may be particularly difficult in the context of AI systems, where risks of manipulation, discrimination, and arbitrary decisions will be certainly occurring’, even when the providers and deployers have prima facie complied with their duty of care as defined by the AI Act. 

In order to overcome these proof-related difficulties, several solutions have been proposed. BEUC, the European Consumer Organisation, has recommended introducing a reversal of the burden of proof[6], in order to allow the consumers to only have to prove the damage they suffered and the involvement of an AI system. A more nuanced approach has been suggested by an expert, aiming at differentiating between AI systems, whether they are high-risk or not, and general-purpose AI systems: providers and deployers of high-risk AI systems would be subjected to ‘truly strict liability’, while SMEs and non-high-risk AI systems should only be subjected to rebuttable presumptions of fault and causality[7]. In the same vein, the European Parliament considered in 2020 that it seemed ‘reasonable to set up a common strict liability regime for (…) high-risk autonomous AI-systems’. As regards other AI systems, the European Parliament also considered that ‘affected persons should nevertheless benefit from a presumption of fault on the part of the operator who should be able to exculpate itself by proving it has abided by its duty of care’[8].

The Commission itself has acknowledged, in its impact assessment report, that ‘the specific characteristics of the AI-system could make the victim’s burden of proof prohibitively difficult or even impossible to meet’, and has evoked different approaches, among which the reversal of the burden of proof. As a sign of its hesitation, the Commission has introduced in the Proposal the possibility to review the directive five years after the end of the transposition period, in particular in order to ‘evaluate the appropriateness of no-fault liability rules for claims against the operators of certain AI systems, as long as not already covered by other Union liability rules, and the need for insurance coverage, while taking into account the effect and impact on the roll-out and uptake of AI systems, especially for SMEs’[9].

II- The withdrawal of the Proposal

On 11 February 2025, the Commission decided to withdraw the Proposal, on the grounds that there was ‘no foreseeable agreement’, and that the Commission would ‘assess whether another proposal should be tabled or another type of approach should be chosen’[10].

This decision caught the European Parliament’s rapporteur on the Proposal, Axel Voss (PPE), by surprise, who stated that the scrapping of the rules would mean ‘legal uncertainty, corporate power imbalances, and a Wild West approach to AI liability that benefits only Big Tech’[11].

On the other hand, the decision of the Commission is reported to have satisfied both the Council and the private sector. In particular, France’s Permanent Representation would have indicated that it saw no reason to impose additional liability requirements on AI providers[12].

How can such a situation be explained?

It is true that the AI liability initiative launched by the Commission on 28 September 2022 was also composed of another Proposal, aiming at updating the Directive on liability for defective products (PLD). The new directive, which now includes software and digital manufacturing files within the definition of product, and expands the notion of compensable damage to include the destruction or corruption of data, came into force on 8 December 2024.

However, the scope of the revised PLD is limited: it only provides compensation for material losses resulting from death, personal injury, damages to property and loss or corruption of data (Article 6 PLD). In particular, damage stemming from a violation of a fundamental right without any material loss is not covered by this directive, but should have been covered by the AI liability directive. The draft AILD aimed at covering ‘national liability claims mainly based on the fault of any person with a view of compensating any type of damage and any type of victim’[13].

The loopholes of the PLD have also been underlined by the complementary impact assessment required by the JURI Committee, to which the file had been attributed in the European Parliament. The study[14], published on 19 September 2024, underlines: ‘However, the PLD presents notable gaps, especially in areas such as protection against discrimination, personality rights, and coverage for professionally used property. It also lacks measures for addressing pure economic loss and sustainability harms, as well as damage caused by consumers, which are contingent on Member State laws. These limitations underscore the necessity for adopting the AILD (…)’.

Thus, in the light of the complementary impact assessment, it appears that the recent adoption of the revised PLD cannot compensate the withdrawal of the proposed AILD. Moreover, as stressed by the first International AI Safety Report, the specific characteristics of general-purpose AI systems make legal liability hard to determine:

‘The fact that general-purpose AI systems can act in ways that were not explicitly programmed or intended by their developers or users raises questions about who should be held liable for resulting harm’[15] .

Conclusion:

Against this background, today the European citizens are left with a ‘fragmented patchwork of 27 different national legal systems’[16], most of them relying on a fault-based regime, which is not able to respond to all the challenges posed by AI systems, and in particular to general-purpose AI systems.

The withdrawal of the proposed AILD is only one element of the Commission’s plan aiming at ‘simplifying rules and effective implementation’[17], which enlists 37 withdrawn proposals in total.

The fact that the final 2025 work programme of the Commission -with the addition of the withdrawal of the AILD- was published just after the AI Act Summit, held in Paris on 10-11 February, may be a simple coincidence. However, it should be noted that the Statement[18] issued after the AI Summit does not refer to the issue of liability nor to the risks of AI systems, except in the context of information.

As observed by Anupriya Datta and Théophane Hartmann in Euractiv, ‘In this context, withdrawing the AI liability directive can be understood as a strategic manoeuvre by the EU to present an image of openness to capital and innovation, to show it prioritises competitiveness and show goodwill to the new US administration’[19].

The final word may not have been spoken, yet. On 18 February, the Members of the European Parliament’s Internal Market and Consumer Protection Committee (IMCO) voted to keep working on liability rules for artificial intelligence products, despite the European Commission’s intention to withdraw the proposal[20].


[1] The opinions expressed in this article are the author’s own and do not necessarily represent the views of the EDPS

[2] International Scientific Report on the Safety of Advanced AI January 2025

[3] COM(2022) 496 final, page 2

[4] COM(2022) 496 final, page 6

[5] EDPS Opinion 42/2023 on the Proposals for two Directives on AI liability rules, 11 October 2023, par. 33

[6] Proposal for an AI liability Directive, BEUC position paper, page 12.

[7] The European AI liability directives – Critique of a half-hearted approach and lessons for the future-Philipp Hacker, page 49.

[8] European Parliament resolution of 20 October 2020 with recommendations to the Commission on a civil liability regime for artificial intelligence (2020/2014(INL)), par. 14 and 20.

[9] Article 5 of the Proposal.

[10] Annexes to the Communication from the Commission to the European Parliament, the Council, the European, Economic and Social Committee and the Committee of the Regions- Commission work programme 2025, page 26.

[11] Euractiv ‘Commission plans to withdraw AI Liability Directive draw mixed reactions’, 12 February 2025.

[12] Ibidem.

[13] COM(2022) 496 final, page 3.

[14] Proposal for a directive on adapting non-contractual civil liability rules to artificial intelligence-Complementary impact assessment.

[15] International Scientific Report on the Safety of Advanced AI, page 179.

[16] Euractiv, ‘Commission plans to withdraw AI Liability Directive draw mixed reactions’, Anupriya Datta, 12 February 2025

[17] Commission work programme 2025, page 11

[18] Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet: ‘We will keep addressing the risks of AI to information integrity and continue the work on AI transparency.’

[19] Euractiv ‘Commission withdraws AI liability directive after Vance attack on regulation’, 11 February 2025

[20] Euronews ‘Lawmakers reject Commission decision to scrap planned AI liability rules’, Cynthia Kroet, 18/02/2025

The Commission proposal of withdrawing the draft AI Liability Directive: a “strategic error” both from an institutional and content perspective? (1)

by Emilio DE CAPITANI

Maybe it is a pure coincidence, but the Commission proposal to withdraw the drat AI liability Directive seems to be the immediate answer to the US Vice-president, J.D. Vance request at the Artificial Intelligence Action Summit in Paris that the EU (and its Member States) should avoid any regulation deemed (by the US) too “aggressive” against the American technology giants. To confirm such suspicion are the Commission Vice President Sefcovic justifications according to which the Commission was withdrawing the text because of the “lack of progress” in the legislative process. This justification is simply unfounded because both the Council and the Parliament are currently actively working on the issue : the Council is debating the reactions of the Member States on the Commission’s proposal (see here and here) and the European Parliament held an hearing on this subject not later than two weeks ago, following which the EP Rapporteur has already announced a draft report in the coming months.

So, the Commission justification for withdrawing the AI Liability Directive proposal because of lack of progress is factually unfounded and, even, legally questionable. (see my other general post Here). Suffice to remember that according to the CJEU judgement withdrawing a legislative proposal may be justified only  ‘where an amendment planned by the Parliament and the Council distorts the proposal for a legislative act in a manner which prevents achievement of the objectives pursued by the proposal and which, therefore, deprives it of its raison d’être, the Commission is entitled to withdraw it’. It may however do so only after having had due regard to Parliament’s and Council’s concerns behind their wish to amend the proposal.” (C- C‑409/13, p.83). Again this seems to be of common sense interpretation. Until the Treaty will not recognize a full right of legislative initiative to the EP and to the Council these institutions may only ask the Commission to submit a legislative proposal (art.225 of the TFEU for the EP).

However, until now no formal amendment depriving the AI Liability proposal of its raison d’être have been tabled by the EP or by the Council so that the co-legislator may well continue to work on the current legislative proposal. Needless to say that, in case of formal withdrawal of the text by the Commission the co-legislators or even only one of them may well bring that institution to the Court for infringement of the principle of conferral and of Institutional Balance.

(continue)

The COE Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law. Is the Council of Europe losing its compass ?

by Emilio DE CAPITANI

When the Committee of Ministers of the Council of Europe decided at the end of 2021 to establish the Committee on Artificial Intelligence (CAI) with the mandate to elaborate a legally binding instrument of a transversal character in the field of artificial intelligence (AI), such initiative created a lot of hopes and expectations. For the first time, an international convention ‘based on the Council of Europe’s standards on human rights, democracy and the rule of law and other relevant international standards’ would regulate activities developed in the area of AI.  

The mandate of the CAI was supposed to further build upon the work of the Ad Hoc Committee on Artificial Intelligence (CAHAI), which adopted its last report in December 2021, presenting  ‘possible elements of a legal framework on artificial intelligence, based on the Council of Europe’s standards on human rights, democracy and the rule of law’. In this document, the CAHAI underlined the need for the future instrument to ‘focus on preventing and/or mitigating risks emanating from applications of AI systems with the potential to interfere with the enjoyment of human rights, the functioning of democracy and the observance of the rule of law, all the while promoting socially beneficial AI applications’. In particular, the CAHAI considered that the instrument should be applicable to the development, design and application of artificial intelligence (AI) systems, ‘irrespective of whether these activities are undertaken by public or private actors’, and that it should be underpinned by a risk-based approach. The risk classification should include ‘a number of categories (e.g., “low risk”, “high risk”, “unacceptable risk”), based on a risk assessment in relation to the enjoyment of human rights, the functioning of democracy and the observance of the rule of law’. According to the CAHAI, the instrument should also include ‘a provision aimed at ensuring the necessary level of human oversight over AI systems and their effects, throughout their lifecycles’.

So, a lot of hopes and expectations: some experts expressed the wish to see this new instrument as a way to complement, at least in the European Union, the future AI Act, seen as a regulation for the digital single market, setting aside the rights of the persons affected by the use of AI  systems[1]. In its opinion of 20/2022 on the Recommendation for a Council Decision authorising the opening of negotiations on behalf of the European Union for this Council of Europe convention, the EDPS considered that it represented ‘an important opportunity to complement the proposed AI Act by strengthening the protection of fundamental rights of all persons affected by AI systems’. The EDPS advocated that the convention should provide ‘clear and strong safeguards for the persons affected by the use of AI systems’.

Alas, those hopes and expectations were quickly dampened by the way the negotiations were organised, and, above all, by the content of the future instrument itself.

1- the organisation of the negotiations: the non-member States leading, the civil society out

The objective to open the future instrument to States which are not members of the Council of Europe was with no doubt an excellent initiative, considering the borderless character of AI, and the need to regulate this technology worldwide. Indeed, as noted by the CAHAI in its above mentioned report ‘The various legal issues raised by the application of AI systems are not specific to the member States of the Council of Europe, but are, due to the many global actors involved and the global effects they engender, transnational in nature’. The CAHAI therefore recommended that the instrument, ‘though obviously based on Council of Europe standards, be drafted in such a way that it facilitates accession by States outside of the region that share the aforementioned standards’. So, yes on a global reach, but provided that the standards of the Council of Europe are fully respected.

However, the conditions under which those non-member States have participated in the negotiations need be looked at a little more: not only have they been part of the drafting group sessions unlike the representatives of the civil society, but it seems that from the start they have played a decisive role in the conduct of negotiations. According to a report published in Euractiv in January 2023[2], the US delegation opposed the publication of the first draft of the Convention (the ‘zero draft’), refusing to disclose its negotiating positions publicly to non-country representatives.

At the same time, the organisation of the negotiations has set aside the civil society groups, who were only allowed to intervene in the plenary sessions of the meetings, while the text was discussed and modified in the drafting sessions. The next and-in principle- last plenary meeting from the 11th to the 14th of March should start with a drafting session and will end with the plenary session, which implies that the civil society representatives will have less than 24 hours to have a look at the revised version of the convention -if they can receive it on time- and make their last comments, assuming that their voices were really heard during the negotiations.

Yet, representatives of the civil society and human rights institutions have done their utmost to play an active part in the negotiations. In an email to the participating States, they recalled that the decision to exclude them from the drafting group went ‘against the examples of good practice from the Council of Europe, the prior practice of the drafting of Convention 108+, and the CoE’s own standards on civil participation in political decision-making”[3]. During the 3rd Plenary meeting of 11-13 January 2023, they insisted on being part of the drafting sessions, but the Chair refused, as indicated in the list of decisions:

‘(…) –Take note of and consider the concerns raised by some Observers regarding the decision taken by the Committee at the occasion of its 2nd Plenary meeting to establish a Drafting Group to prepare the draft [Framework] Convention, composed of potential Parties to the [Framework] Convention and reporting to the Plenary.

– Not to revise the aforesaid decision, while underlining the need to ensure an inclusive and transparent negotiation process involving all Members, Participants and Observers and endorsing the Chair’s proposal for working methods in this regard’.[4]

Despite this commitment, the need of an ‘inclusive and transparent negotiation process’ has not been ensured in the light of the civil society statement of the 4th of July 2023, where again the authors ‘deeply regret(ted) that the negotiating States have chosen to exclude both civil society observers and Council of Europe member participants from the formal and informal meetings of the drafting group of the Convention. This undermines the transparency and accountability of the Council of Europe and is contrary to the established Council of Europe practice and the Committee on AI (CAI) own Terms of Reference which instructs the CAI to “contribute[…] to strengthening the role and meaningful participation of civil society in its work”.’[5]

The influence of non-member States has not been limited to the organisation of meetings. As detailed below, the American and Canadian delegations delegations, among others, threw their full weight behind the choice of systematically watering down the substance of the Convention.

2- A convention with no specific rights and very limited obligations

How should the mandate of the CAI be understood? According to the terms of reference, the Committee is instructed to ‘establish an international negotiation process and conduct work to finalise an appropriate legal framework on the development, design, use and decommissioning of artificial intelligence, based on the Council of Europe’s standards on human rights, democracy and the rule of law and other relevant international standards, and conducive to innovation, which can be composed of a binding legal instrument of a transversal character, including notably general common principles (…)’[6].

The objective of including in the convention ‘general common principles’ has been interpreted by the Chair literally, who considered that ‘the AI Convention will offer an underlying baseline of principles in how to handle the technology, on top of which individual governments can then build their own legislation to meet their own specific needs’[7]. Indeed, the last publicly available version -dated 18 December 2023- of the draft Convention only refers to ‘principles’ and not to specific rights[8], even those already existing in the framework of the Council of Europe and beyond. In the context of AI, though, one could have hoped the recognition of certain rights, as the right to human oversight and the right to explanation for AI based decisions.

Such a choice has been criticized by the civil society‘s representatives. In a public statement of the 4th of July 2023, they recalled that ‘while including general common principles for AI regulation as indicated in the CAI Terms of Reference, the Convention should respect the rights established by other Conventions and not reformulate them as mere principles’[9].

Unfortunately, the Convention, at least in the version of the 18th of December 2023, does not even expressly include the right to privacy and the right to the protection of personal data. Yet, if data are, as the Chair himself referred to, ‘the oil of the XX1st century’[10], the need to protect our rights in this area is critical.

If one compares the successive versions of the Convention which are publicly accessible, from the zero draft[11], to the version of the 18th of December, one can only deplore the constant watering down of its content. What about ‘prohibited artificial intelligence practices’ referred to in Article 14 of the zero draft? What about the definitions, which included in the zero draft the notion of ‘artificial intelligence subject’, defined as ‘any natural or legal person whose human rights and fundamental freedoms, legal rights or interests are impacted by decisions made or substantially informed by the application use  of an artificial intelligence system’? What about a clear presentation of the risk-based approach, with a differentiation of measures to be applied in respect of artificial intelligence systems posing significant and unacceptable levels of risk (see articles 12 and 13 of the zero draft)?

Moreover, in the version of the 18th of December 2023, a number of obligations in principle imposed on Parties might become simple obligations of means, since the possible -or already accepted- wording would be that each party should ‘seek to ensure’ that adequate measures are in place. It is in particular the case in the article dedicated to the ‘integrity of democratic processes and respect for rule of law’, as well as in the article on ‘accountability and responsibility’ and even in the article on procedural safeguards, when persons are interacting with an artificial intelligence system without knowing it.

According to an article published in Euractiv on 31 Jan 2024 and updated on 15 Feb 2024, even the version of the 18th of December 2023 seems to have been watered down: ‘Entire provisions, such as protecting health and the environment, measures promoting trust in AI systems, and the requirement to provide human oversight for AI-driven decisions affecting people’s human rights, have been scrapped’[12].

3- The worse to come?

One crucial element of the Convention still needs to be discussed: its scope. Since the beginning of the negotiations, the USA and Canada, but also Japan and Israel, none of them members of the Council of Europe, have clearly indicated their wish to limit the scope of the instrument to activities within the lifecycle of artificial intelligence systems only undertaken by public authorities[13]. Moreover, national security and defence should also be out of the scope of the convention.  The version of the 18th of December includes several wordings regarding the exemption of national security, which reflect different levels of exemption.

The issue of the scope has lead the representatives of the civil society to draft an open letter[14], signed by an impressive number of organisations calling on the EU and the State Parties negotiating the text of the Convention to equally cover the public and private sectors and to unequivocally reject blanket exemptions regarding national security and defence.

Today no one knows what the result of the last round of negotiations will be: it seems that the EU is determined to maintain its position in favour of the inclusion of the private sector in the scope of the Convention, while the Americans and Canadians might use the signature of the Convention as blackmail to ensure the exclusion of the private sector.

4- Who gains?

From the Council of Europe perspective, which is an organisation founded on the values of human rights, democracy and the rule of law. the first question that comes to mind is what are the expected results of the ongoing negotiations. Can the obsession to see the Americans sign the Convention justify such a weakened text, even with the private sector in its scope? What would be the gain for the Council of Europe and its member States, to accept a Convention which looks like a simple Declaration, not very far in fact from the Organisation for Economic Co-operation and Development’s Principles on AI[15]?

At this stage, it seems that neither the Americans nor the Canadians are ready to sign the Convention with the inclusion of the private sector, even if an opt-out clause were inserted in the text. The gamble of the Chair and the Secretariat to keep these two observer States on board at the price of excessive compromises might be lost at the end of the day. One should not forget that these States do not have voting rights in the Committee of Ministers.

The second question that comes to mind is why the Chair and the Secretariat of the CAI and, above them, those who lead the Council of Europe have made such a choice. Does it have a link with internal decisions to be taken in the next future, as regards the post of the General Secretary of the organisation, as well as the post of the Director General of Human Rights and Rule of Law? Does the nationality of the Chair have a role to play in this game? In any case, the future Convention might look like an empty shell, which might have more adverse effects than it seems prima facie, by legitimizing practices around the world which would be considered incompatible with the European standards.

NOTES


[1] See in particular ‘The Council of Europe’s road towards an AI Convention: taking stock’ by Peggy Valcke and Victoria Hendrickx, 9 February 2023: ‘Whereas the AI Act focuses on the digital single market and does not create new rights for individuals, the Convention might fill these gaps by being the first legally binding treaty on AI that focuses on democracy, human rights and the rule of law’. https://www.law.kuleuven.be/citip/blog/the-council-of-europes-road-towards-an-ai-convention-taking-stock/

[2] https://www.euractiv.com/section/digital/news/us-obtains-exclusion-of-ngos-from-drafting-ai-treaty/

[3] same article

[4] https://rm.coe.int/cai-2023-03-list-of-decisions/1680a9cc4f

[5] https://ecnl.org/sites/default/files/2023-07/CSO-COE-Statement_07042023_Website.pdf

[6] https://rm.coe.int/terms-of-reference-of-the-committee-on-artificial-intelligence-cai-/1680ade00f

[7] https://www.politico.eu/newsletter/digital-bridge/one-treaty-to-rule-ai-global-politico-transatlantic-data-deal/

[8] with the exception of ‘rights of persons with disabilities and of children’ in Article 18

[9] https://ecnl.org/sites/default/files/2023-07/CSO-COE-Statement_07042023_Website.pdf

[10] https://www.linkedin.com/pulse/data-oil-21st-century-ai-systems-engines-digital-thomas-schneider/

[11] https://www.statewatch.org/news/2023/january/council-of-europe-convention-on-artificial-intelligence-zero-draft-and-member-state-submissions/

[12] https://www.euractiv.com/section/artificial-intelligence/news/tug-of-war-continues-on-international-ai-treaty-as-text-gets-softened-further/

[13] same article

[14] https://docs.google.com/document/d/19pwQg0r7g5Dm6_OlRvTAgBPGXaufZrNW/edit

[15] https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449