European politicians are currently traveling around the world to mediate in the conflict between Russia and Ukraine. Emmanuel Macron is visiting Vladimir Putin in Moscow today, Annalena Baerbock is traveling to Kiev, and Olaf Scholz is flying to Washington for his inaugural visit to Joe Biden.
There are increasingly loud calls from the USA for Nord Stream 2 not to go into operation to put Russia under pressure. Meanwhile, Olaf Scholz hinted that a halt to the gas pipeline could not be ruled out in the event of an escalation of the conflict.
In any case, Europe is keen to reduce its dependence on Russian natural gas. This is the second reason for numerous trips abroad by European politicians these days. Energy Commissioner Kadri Simson and EU Foreign Affairs Commissioner Josep Borrell are also in Washington today. They will join Secretary of State Antony Blinken and Secretary of Energy Jennifer Granholm at the EU-US Energy Council to intensify talks on future cooperation on energy security and the joint commitment to accelerate the energy transition toward climate neutrality.
Just this Friday, Simson visited Baku to promote higher gas exports from Azerbaijan to Europe. Read more about Simson’s visit in the News.
However, new energy supply contracts between Europe and other countries could also bring new problems. If the EU secures LNG from China, emerging countries in Southeast Asia could find themselves in energy trouble instead, explains Andreas Goldthau in an interview with Charlotte Wirth. Poorer countries would not be able to keep up with the prices Europe could pay. The problem would shift to others, according to the energy expert.
The EU Parliament is currently negotiating the regulation of artificial intelligence. The Council is already much further along. Both the Slovenian and French presidencies have already put forward their own proposals. Jasmin Kohl points out the lines of conflict between the member states and in the Parliament. There is disagreement above all on which systems should be classified as high-risk AI systems.
Mr. Goldthau, how serious do you think the energy crisis in Europe is?
We are currently discussing energy prices and security of supply. Here, the situation is undoubtedly serious. But if war really breaks out in Ukraine, we have a scenario that goes far beyond that. It would be about the stability of a country, about humanitarian crises, people will have to flee to the West and be supplied. In Germany, we are looking very narrowly at whether we will still have our natural gas tomorrow and whether Nord Stream 2 can be maintained. Our problem in the event of a war is not the gas supply but the war that will then rage in the middle of Europe.
Could Russia even economically afford a war and a halt in gas supplies?
Seven or eight percent of Russia’s state budget comes from natural gas revenues, 40 percent from oil. If gas revenues were to disappear, that would be dramatic. But it would not mean state bankruptcy. However, Gazprom has another role in Russia; it is to contribute to the country’s economic development. Entire cities would no longer exist if Gazprom did not build soccer stadiums, operate public swimming pools and other infrastructure there. Gazprom has to refinance this development contract, and it does this through exports. One-third of its gas is exported, but this is how the company generates 100 percent of its revenues.
So Gazprom is vulnerable here?
On the one hand, yes. On the other hand, the oil price would probably go through the roof if Russia invades Ukraine, and so would Russian oil revenues. If I were a Russian portfolio manager in government service and knew whether Russia would invade, I would place options and leverage the effect. The losses from the lack of gas exports would thus refinance themselves. And the Russians are sitting on a big sovereign wealth fund. They are no longer as vulnerable as they were when Crimea was annexed.
Do economic sanctions against Russia make more sense?
Sanctions are unavoidable when it comes to acts of war. But we have to consider what this means. For example, suppose we shut down Swift for the Russians: How will we pay for our gas then? That’s what we need to prepare for.
The EU Commission and the USA are preparing Europe for a supply crisis. How dramatic would it be if Russia actually cut gas supplies?
That depends on how long the situation will last. It would probably be a manageable disruption if we were dealing with a brief act of war in eastern Ukraine. We could certainly compensate for that in the short term; we’re already importing more liquefied gas than ever before. Price-wise, though, it wouldn’t be pretty. We are in tough competition with Asia.
And in the event of a medium-term supply interruption?
There are still capacities on the LNG market that have not been exhausted. However, these are not particularly large.
From where could the liquid gas come?
We already see LNG flows change. Chinese suppliers, for example, are starting to send LNG to Europe – that’s new. However, this also means that we are pricing other countries out of the market. These are then emerging countries, for example, in Southeast Asia, which can no longer pay called prices for the LNG. Then the LNG comes to us, but we shift the problem to others. The shortage on the natural gas market is global, not only here.
Europe could make greater use of other fuels.
In Germany, many coal-fired power plants are not operating at full capacity, and others have already been shut down. These could be brought back on stream. This is a strategic reserve that can be used in the short term to cover part of the electricity or heat load that would otherwise be covered by gas.
In terms of climate policy, however, this is problematic.
That would certainly be a decision made in the case of force majeure. But by bringing the coal plants back, of course, you’re pushing the whole decarbonization back. Investment decisions would possibly be affected. So that wouldn’t be a simple supply bridge; it would potentially have systemic implications.
Will the situation improve in summer?
We have seasonal fluctuations in demand: You also need gas in summer, but for other things. If the shortage persists and Russian gas fails into next winter, we could adjust energy demand, for example, by shutting down parts of the industry or sending people into home office. The very last option would be to turn down the heat supply to private households, but that would be an emergency measure.
What would be a long-term scenario?
The amount of gas that the Russians are sending to the West cannot be intercepted in the long term via LNG. We would have to come up with something else.
Ms. von der Leyen is now negotiating with countries like the USA and Qatar. What do you think that will achieve?
Nothing in the short term. Neither Ms. von der Leyen nor the Emir of Qatar nor Joe Biden can turn the supply screw. In the US, the gas industry is privately organized. Even if Biden wanted to send more LNG to Europe, he couldn’t divert cargo ships and invent production capacity that doesn’t exist. Much of the available LNG is also tied up in long-term contracts, for example, with Southeast Asia.
And what if Europe now also concludes long-term contracts?
It is difficult to conclude long-term contracts with European importers. This is because certain clauses may no longer be included, as they are incompatible with European energy law. These are the so-called territoriality clauses, which prohibit resale. They were abolished so that gas arriving from different directions could be traded within the EU. This makes sense from a competition point of view. However, it can have consequences on the supply side. When the competition case against Qatar Petroleum came up in 2018, the Qataris, therefore, turned away from Europe and toward Southeast Asia. So the EU Commission would have to end the competition case first if the Qataris are to reconsider.
The climate issue seems to be fading into the background at the moment. Is decarbonization still being considered in the current crisis?
After all, this is primarily a price crisis. That’s where firefighting is needed. You have to cushion the effect on the families and economic sectors affected, and that is already happening. The bigger question is: What lessons do we learn from this? On the one hand, strong voices say we should not blindly focus on decarbonization with a view to energy prices, even as deindustrialization threatens. Others say: now more than ever! Decarbonization means turning away from fossil fuels, which are causing problems.
What do you say?
We need to bring more resilience into the system so that swings in demand or price can be cushioned. But to rethink the whole decarbonization course would be to create an infrastructure that would eventually be on the books as stranded assets. And you would miss the important impetus for Europe’s economy toward a green growth model.
While negotiations on the artificial intelligence regulation have just begun in the European Parliament (Europe.Table reported), the Council has already made significant progress. In mid-January, the French Council Presidency presented its first compromise proposal. It concerns the requirements for high-risk AI systems (Chapter 2, Articles 8-15 and Annex 4).
In Article 10 (3) on data and data governance, a significant change to the Commission’s proposal can be found: In it, the Commission had required that training, validation, and test data sets for high-risk AI systems have to be complete and error-free. However, many member states found this requirement difficult to meet. The French Council Presidency took this hint on board and advocated that these data sets only need to be error-free and complete “within the bounds of possibility”.
With regard to the technical documentation of a high-risk AI system (Article 11), the French compromise proposal provides for more flexibility for start-ups and SMEs. Accordingly, these companies would not have to list all the detailed information listed in Annex IV in the technical documentation of the systems. It would be sufficient if they provide “equivalent” information that has the same objective: Demonstrate that the high-risk AI system meets the requirements of Chapter 2, such as the requirement for recordkeeping and the establishment of a risk management system. Competent authorities are to determine whether the information is indeed equivalent.
The French Presidency, on the other hand, is calling for a more ambitious approach to transparency and the provision of information to users (Article 13): The instructions for the use of a high-risk AI system should also include a description of the system mechanism that allows users to properly record, store, and interpret logs.
Shortly before the handover to France, the Slovenian Council Presidency had published its progress report at the end of November as well as a first compromise proposal for Articles 1-7 and Annexes 1-4 of the regulation. In mid-January, the member states’ representatives discussed it for the first time at working group level. On Monday, there was a further exchange. Reactions have been mostly positive, but many delegations have not yet found time to go through the entire compromise proposal, a Council official reports. Deeper discussions on these articles at the working group level are therefore still pending.
“Contexte” was the first to get access last week to excerpts (Document 1 and Document 2) of the comments of 20 member states on the first 29 articles of the AI regulation. The Slovenian Council Presidency had requested these in order to prepare its compromise proposal. The comments from mid-November give first insights into the member states’ positions.
The German position is not apparent from the documents. Claudia Dörr-Voß, State Secretary for Energy and Economic Affairs under the old German government, had said at an informal Council video conference in mid-October (Europe.Table reported) that the government’s internal review of the Commission’s proposal for AI regulation was not yet complete due to its complexity and shared these key points. The German government welcomes the horizontal and risk-based regulatory approach, sees a need for improvement in the list of prohibited AI applications as well as in the list of high-risk AI applications, and considers the support of SMEs and start-ups within the regulation to be very important.
It is unclear to what extent the new German government will continue to adhere to this position or reconsider it. When asked by Europe.Table, the Federal Ministry for Economic Affairs and Climate Action (BMWK) did not want to comment on this and also left unanswered whether the examination of the Commission’s proposal for the AI regulation was still ongoing.
In its compromise draft, the Slovenian Council Presidency also advocated classifying AI systems that act to control digital infrastructure or as its security components as high-risk AI systems (Annex 3). AI systems that control emissions and pollution should also be included in this category. The demand of Susana Solís Pérez (Renew), rapporteur of the EU Parliament’s Committee on Environment (ENVI), goes even further on this point: She wants to classify all AI systems in the field of environment as high-risk AI systems.
However, AI systems used by law enforcement agencies for crime analysis should not be classified as high-risk AI systems. The Czech Republic is generally very skeptical about the classification because it fundamentally questions that high-risk AI systems can cause harm. Thus, if the Commission does not provide evidence of this, the references to these systems should be removed entirely from the text of the law “as they are unjustified”, the country commented in November.
Portugal favors adding healthcare to the list of high-risk AI systems. Thus, AI systems that help diagnose, check vital signs, or create medical treatment plans would also be classified as high-risk AI systems. France proposed in November to adapt the list of high-risk AI systems through implementing acts rather than delegated acts, as envisioned by the Commission. Through the comitology procedure, member states would thus play a significant role in the decision-making process. In addition, the adaptation of the list is not to take place at any time as required but only at fixed intervals. The intervals at which this is to take place remain open.
According to the compromise draft of the Slovenian Council Presidency, so-called social scoring (assessment of social behavior) is to be prohibited not only for public but also for private actors. In contrast, automated real-time facial recognition in public spaces for law enforcement is to be allowed to be used not only by law enforcement authorities themselves but also “by other actors on behalf of law enforcement authorities”. Examples of use include preventing terrorist attacks or searching for missing children. This holds potential for dispute with the European Parliament because the rapporteur of the Committee on Internal Market and Consumer Protection (IMCO), Brando Benifei (S&D), wants to fundamentally ban the use of the controversial practice – even for the purpose of law enforcement.
In addition, the compromise draft provides for extending automated real-time facial recognition to the protection of critical infrastructure. Spain favored allowing the technology to be used at events that pose a risk to public order, such as sports competitions or summits. French MEP Gwendoline Delbos-Corfield (Greens/EFA) is convinced that France will join this demand (Europe.Table reported). For the 2024 Olympics in Paris, the Île-de-France region would already explore the possibility of using biometric mass surveillance.
The Commission had proposed that a judicial authority should authorize any use of automated real-time facial recognition. For emergencies such as terrorist attacks, however, ex-post approval should be sufficient. This is contradicted by the compromise proposal of the Slovenian Council Presidency: The authorization is to be applied for during the deployment. If it is rejected, the deployment must be stopped immediately.
With a clear reference to national competencies, the Slovenian Council Presidency has argued that national security should be removed from the scope of the AI regulation. This would continue to fall under the sole responsibility of the individual member states. AI systems that are used solely for research and development purposes are also not to be affected by the law.
In order to prevent traditional software programs from falling under the definition of AI systems (Article 3) and thus the scope of the regulation, the Slovenian Council Presidency has narrowed the definition accordingly. Roberto Viola, Director General of the DG Connect, had most recently defended the definition in the Commission proposal against similar criticism during the first joint debate of the lead committees in the European Parliament IMCO and LIBE (Committee on Civil Liberties, Justice and Home Affairs): It was broad enough to include the various AI systems, but also sufficiently narrow to exclude simple systems that only process data.
The French Council Presidency plans to present a new compromise proposal on the obligations of providers and users of high-risk AI systems and other stakeholders (Articles 16-29) of the regulation on February 10. Another compromise proposal on Articles 30-85 and Annexes 5-9 is currently scheduled for February 22.
The lead committees in the European Parliament, IMCO and LIBE, plan to hold a first hearing on the AI regulation with experts and scientists on March 16. Rapporteurs Brando Benifei (S&D) and Dragoş Tudorache (Renew) plan to draft their joint report by April 5.
Like after every good crisis, a sea change is currently taking place in EU economic governance. While much of the public reform discussions are focusing on fiscal rules and the Stability and Growth Pact (SGP), a more silent revolution is also going in EU economic policy coordination. In a new paper, we argue that this revolution will change EU economic governance for good. And that its current form needs a serious rethink.
It is a less known feature of the EU’s new €800 billion joint recovery fund that it has been tightly wed with the so-called European Semester. The Semester is the EU’s main tool to steer member states towards coordinating their national fiscal, labor market, or social policies. For that, the Commission monitors national economies and hands-out annual reform recommendations (CSRs) to the capitals.
Generally, the Semester has a bad reputation. It is perceived as overly bureaucratic, the process lacks sticks and carrots, member states have little incentives to follow the Commission’s advice and compliance has been low and declining. With the new Recovery and Resilience Fund (RRF), this changes.
To tap into the recovery money, capitals must submit detailed national recovery and resilience plans (NRRPs) to the Commission. These plans must not just outline how the new European money is spent but also include reform agendas that address “a substantial subset” of the CSRs issued under the Semester. The money then comes in tranches. Member states only receive the next payment if the Commission is satisfied with their investment and reform progress.
So, until the RRF ends in 2026, the Commission can now in theory hold back billions in loans and grants if member states do not follow agreed reform plans. This adds a lot of teeth to EU economic policy coordination and increases the leverage of the Commission. It will also have a lasting impact on EU economic governance. Thus, getting this right is important. We think that this requires three things.
First, the Commission should scale back recommendations under the regular Semester. By linking the fulfillment of those CSRs included in NRRPs to money, the RRF effectively makes some CSRs much more binding than others. In the next five years, all eyes will be on the national recovery and resilience plans and the implementation of the RRF. Long additional lists of CSRs outside the RRF are bound to be ignored. The Commission should therefore scale down its ambitions about what can be achieved with the Semester outside the RRF.
Second, the RRF introduces asymmetric leverages into EU economic coordination. Some member states like Hungary, Greece, Spain, or Italy are eligible to large sums of loans and grants under the RRF. These countries had to submit ambitious reform plans and will come under a lot of pressure from Brussels in the coming years.
Other countries like Germany or the Netherlands receive less and can ignore the Commission’s wishes more easily. Politically and economically, this is problematic. The Commission, thus, should not ask too much from countries in dire need of RRF funds while leaving economically-well off ones off the hook on key reforms. It should make sure that money flows only in return for progress – no matter the states or sums involved. And it should make full use of the Macroeconomic Imbalances Procedure (MIP) to keep countries that receive little RRF money and have more fiscal space like Germany on the radar.
Third, national reform and investment plans will need to be adapted. Effectively, the NRRPs lay out more or less detailed milestones and targets for each and every investment and reform that member states have committed to undertake by 2026. A lot will happen between now and 2026. Investments can fail, European priorities can evolve, and national governments can change.
If so, member states and the Commission will need to bargain whether and how to adopt plans. To turn the new Semester into an effective coordination tool, the Commission should not insist on following original plans by the letter. Instead, it should use instances of renegotiating parts of the plans as an opportunity to constantly update areas where national and European priorities overlap.
The new marriage between the RRF and the European Semester will shape EU economic governance long beyond its official expiration date in 2026. The Commission has just re-opened a public review of the EU’s economic governance framework.
The link between the RRF and the Semester means that tinkering at the margins of the old system is no longer an option. Any debate on the future of the Semester will now take place against the backdrop of the RRF experience. And any big future fiscal capacity at the EU level will almost certainly come with a Semester-style reform leg. The next years will determine what direction this discussion will take but in any case, it will change the EU’s economic landscape for good.
European politicians are currently traveling around the world to mediate in the conflict between Russia and Ukraine. Emmanuel Macron is visiting Vladimir Putin in Moscow today, Annalena Baerbock is traveling to Kiev, and Olaf Scholz is flying to Washington for his inaugural visit to Joe Biden.
There are increasingly loud calls from the USA for Nord Stream 2 not to go into operation to put Russia under pressure. Meanwhile, Olaf Scholz hinted that a halt to the gas pipeline could not be ruled out in the event of an escalation of the conflict.
In any case, Europe is keen to reduce its dependence on Russian natural gas. This is the second reason for numerous trips abroad by European politicians these days. Energy Commissioner Kadri Simson and EU Foreign Affairs Commissioner Josep Borrell are also in Washington today. They will join Secretary of State Antony Blinken and Secretary of Energy Jennifer Granholm at the EU-US Energy Council to intensify talks on future cooperation on energy security and the joint commitment to accelerate the energy transition toward climate neutrality.
Just this Friday, Simson visited Baku to promote higher gas exports from Azerbaijan to Europe. Read more about Simson’s visit in the News.
However, new energy supply contracts between Europe and other countries could also bring new problems. If the EU secures LNG from China, emerging countries in Southeast Asia could find themselves in energy trouble instead, explains Andreas Goldthau in an interview with Charlotte Wirth. Poorer countries would not be able to keep up with the prices Europe could pay. The problem would shift to others, according to the energy expert.
The EU Parliament is currently negotiating the regulation of artificial intelligence. The Council is already much further along. Both the Slovenian and French presidencies have already put forward their own proposals. Jasmin Kohl points out the lines of conflict between the member states and in the Parliament. There is disagreement above all on which systems should be classified as high-risk AI systems.
Mr. Goldthau, how serious do you think the energy crisis in Europe is?
We are currently discussing energy prices and security of supply. Here, the situation is undoubtedly serious. But if war really breaks out in Ukraine, we have a scenario that goes far beyond that. It would be about the stability of a country, about humanitarian crises, people will have to flee to the West and be supplied. In Germany, we are looking very narrowly at whether we will still have our natural gas tomorrow and whether Nord Stream 2 can be maintained. Our problem in the event of a war is not the gas supply but the war that will then rage in the middle of Europe.
Could Russia even economically afford a war and a halt in gas supplies?
Seven or eight percent of Russia’s state budget comes from natural gas revenues, 40 percent from oil. If gas revenues were to disappear, that would be dramatic. But it would not mean state bankruptcy. However, Gazprom has another role in Russia; it is to contribute to the country’s economic development. Entire cities would no longer exist if Gazprom did not build soccer stadiums, operate public swimming pools and other infrastructure there. Gazprom has to refinance this development contract, and it does this through exports. One-third of its gas is exported, but this is how the company generates 100 percent of its revenues.
So Gazprom is vulnerable here?
On the one hand, yes. On the other hand, the oil price would probably go through the roof if Russia invades Ukraine, and so would Russian oil revenues. If I were a Russian portfolio manager in government service and knew whether Russia would invade, I would place options and leverage the effect. The losses from the lack of gas exports would thus refinance themselves. And the Russians are sitting on a big sovereign wealth fund. They are no longer as vulnerable as they were when Crimea was annexed.
Do economic sanctions against Russia make more sense?
Sanctions are unavoidable when it comes to acts of war. But we have to consider what this means. For example, suppose we shut down Swift for the Russians: How will we pay for our gas then? That’s what we need to prepare for.
The EU Commission and the USA are preparing Europe for a supply crisis. How dramatic would it be if Russia actually cut gas supplies?
That depends on how long the situation will last. It would probably be a manageable disruption if we were dealing with a brief act of war in eastern Ukraine. We could certainly compensate for that in the short term; we’re already importing more liquefied gas than ever before. Price-wise, though, it wouldn’t be pretty. We are in tough competition with Asia.
And in the event of a medium-term supply interruption?
There are still capacities on the LNG market that have not been exhausted. However, these are not particularly large.
From where could the liquid gas come?
We already see LNG flows change. Chinese suppliers, for example, are starting to send LNG to Europe – that’s new. However, this also means that we are pricing other countries out of the market. These are then emerging countries, for example, in Southeast Asia, which can no longer pay called prices for the LNG. Then the LNG comes to us, but we shift the problem to others. The shortage on the natural gas market is global, not only here.
Europe could make greater use of other fuels.
In Germany, many coal-fired power plants are not operating at full capacity, and others have already been shut down. These could be brought back on stream. This is a strategic reserve that can be used in the short term to cover part of the electricity or heat load that would otherwise be covered by gas.
In terms of climate policy, however, this is problematic.
That would certainly be a decision made in the case of force majeure. But by bringing the coal plants back, of course, you’re pushing the whole decarbonization back. Investment decisions would possibly be affected. So that wouldn’t be a simple supply bridge; it would potentially have systemic implications.
Will the situation improve in summer?
We have seasonal fluctuations in demand: You also need gas in summer, but for other things. If the shortage persists and Russian gas fails into next winter, we could adjust energy demand, for example, by shutting down parts of the industry or sending people into home office. The very last option would be to turn down the heat supply to private households, but that would be an emergency measure.
What would be a long-term scenario?
The amount of gas that the Russians are sending to the West cannot be intercepted in the long term via LNG. We would have to come up with something else.
Ms. von der Leyen is now negotiating with countries like the USA and Qatar. What do you think that will achieve?
Nothing in the short term. Neither Ms. von der Leyen nor the Emir of Qatar nor Joe Biden can turn the supply screw. In the US, the gas industry is privately organized. Even if Biden wanted to send more LNG to Europe, he couldn’t divert cargo ships and invent production capacity that doesn’t exist. Much of the available LNG is also tied up in long-term contracts, for example, with Southeast Asia.
And what if Europe now also concludes long-term contracts?
It is difficult to conclude long-term contracts with European importers. This is because certain clauses may no longer be included, as they are incompatible with European energy law. These are the so-called territoriality clauses, which prohibit resale. They were abolished so that gas arriving from different directions could be traded within the EU. This makes sense from a competition point of view. However, it can have consequences on the supply side. When the competition case against Qatar Petroleum came up in 2018, the Qataris, therefore, turned away from Europe and toward Southeast Asia. So the EU Commission would have to end the competition case first if the Qataris are to reconsider.
The climate issue seems to be fading into the background at the moment. Is decarbonization still being considered in the current crisis?
After all, this is primarily a price crisis. That’s where firefighting is needed. You have to cushion the effect on the families and economic sectors affected, and that is already happening. The bigger question is: What lessons do we learn from this? On the one hand, strong voices say we should not blindly focus on decarbonization with a view to energy prices, even as deindustrialization threatens. Others say: now more than ever! Decarbonization means turning away from fossil fuels, which are causing problems.
What do you say?
We need to bring more resilience into the system so that swings in demand or price can be cushioned. But to rethink the whole decarbonization course would be to create an infrastructure that would eventually be on the books as stranded assets. And you would miss the important impetus for Europe’s economy toward a green growth model.
While negotiations on the artificial intelligence regulation have just begun in the European Parliament (Europe.Table reported), the Council has already made significant progress. In mid-January, the French Council Presidency presented its first compromise proposal. It concerns the requirements for high-risk AI systems (Chapter 2, Articles 8-15 and Annex 4).
In Article 10 (3) on data and data governance, a significant change to the Commission’s proposal can be found: In it, the Commission had required that training, validation, and test data sets for high-risk AI systems have to be complete and error-free. However, many member states found this requirement difficult to meet. The French Council Presidency took this hint on board and advocated that these data sets only need to be error-free and complete “within the bounds of possibility”.
With regard to the technical documentation of a high-risk AI system (Article 11), the French compromise proposal provides for more flexibility for start-ups and SMEs. Accordingly, these companies would not have to list all the detailed information listed in Annex IV in the technical documentation of the systems. It would be sufficient if they provide “equivalent” information that has the same objective: Demonstrate that the high-risk AI system meets the requirements of Chapter 2, such as the requirement for recordkeeping and the establishment of a risk management system. Competent authorities are to determine whether the information is indeed equivalent.
The French Presidency, on the other hand, is calling for a more ambitious approach to transparency and the provision of information to users (Article 13): The instructions for the use of a high-risk AI system should also include a description of the system mechanism that allows users to properly record, store, and interpret logs.
Shortly before the handover to France, the Slovenian Council Presidency had published its progress report at the end of November as well as a first compromise proposal for Articles 1-7 and Annexes 1-4 of the regulation. In mid-January, the member states’ representatives discussed it for the first time at working group level. On Monday, there was a further exchange. Reactions have been mostly positive, but many delegations have not yet found time to go through the entire compromise proposal, a Council official reports. Deeper discussions on these articles at the working group level are therefore still pending.
“Contexte” was the first to get access last week to excerpts (Document 1 and Document 2) of the comments of 20 member states on the first 29 articles of the AI regulation. The Slovenian Council Presidency had requested these in order to prepare its compromise proposal. The comments from mid-November give first insights into the member states’ positions.
The German position is not apparent from the documents. Claudia Dörr-Voß, State Secretary for Energy and Economic Affairs under the old German government, had said at an informal Council video conference in mid-October (Europe.Table reported) that the government’s internal review of the Commission’s proposal for AI regulation was not yet complete due to its complexity and shared these key points. The German government welcomes the horizontal and risk-based regulatory approach, sees a need for improvement in the list of prohibited AI applications as well as in the list of high-risk AI applications, and considers the support of SMEs and start-ups within the regulation to be very important.
It is unclear to what extent the new German government will continue to adhere to this position or reconsider it. When asked by Europe.Table, the Federal Ministry for Economic Affairs and Climate Action (BMWK) did not want to comment on this and also left unanswered whether the examination of the Commission’s proposal for the AI regulation was still ongoing.
In its compromise draft, the Slovenian Council Presidency also advocated classifying AI systems that act to control digital infrastructure or as its security components as high-risk AI systems (Annex 3). AI systems that control emissions and pollution should also be included in this category. The demand of Susana Solís Pérez (Renew), rapporteur of the EU Parliament’s Committee on Environment (ENVI), goes even further on this point: She wants to classify all AI systems in the field of environment as high-risk AI systems.
However, AI systems used by law enforcement agencies for crime analysis should not be classified as high-risk AI systems. The Czech Republic is generally very skeptical about the classification because it fundamentally questions that high-risk AI systems can cause harm. Thus, if the Commission does not provide evidence of this, the references to these systems should be removed entirely from the text of the law “as they are unjustified”, the country commented in November.
Portugal favors adding healthcare to the list of high-risk AI systems. Thus, AI systems that help diagnose, check vital signs, or create medical treatment plans would also be classified as high-risk AI systems. France proposed in November to adapt the list of high-risk AI systems through implementing acts rather than delegated acts, as envisioned by the Commission. Through the comitology procedure, member states would thus play a significant role in the decision-making process. In addition, the adaptation of the list is not to take place at any time as required but only at fixed intervals. The intervals at which this is to take place remain open.
According to the compromise draft of the Slovenian Council Presidency, so-called social scoring (assessment of social behavior) is to be prohibited not only for public but also for private actors. In contrast, automated real-time facial recognition in public spaces for law enforcement is to be allowed to be used not only by law enforcement authorities themselves but also “by other actors on behalf of law enforcement authorities”. Examples of use include preventing terrorist attacks or searching for missing children. This holds potential for dispute with the European Parliament because the rapporteur of the Committee on Internal Market and Consumer Protection (IMCO), Brando Benifei (S&D), wants to fundamentally ban the use of the controversial practice – even for the purpose of law enforcement.
In addition, the compromise draft provides for extending automated real-time facial recognition to the protection of critical infrastructure. Spain favored allowing the technology to be used at events that pose a risk to public order, such as sports competitions or summits. French MEP Gwendoline Delbos-Corfield (Greens/EFA) is convinced that France will join this demand (Europe.Table reported). For the 2024 Olympics in Paris, the Île-de-France region would already explore the possibility of using biometric mass surveillance.
The Commission had proposed that a judicial authority should authorize any use of automated real-time facial recognition. For emergencies such as terrorist attacks, however, ex-post approval should be sufficient. This is contradicted by the compromise proposal of the Slovenian Council Presidency: The authorization is to be applied for during the deployment. If it is rejected, the deployment must be stopped immediately.
With a clear reference to national competencies, the Slovenian Council Presidency has argued that national security should be removed from the scope of the AI regulation. This would continue to fall under the sole responsibility of the individual member states. AI systems that are used solely for research and development purposes are also not to be affected by the law.
In order to prevent traditional software programs from falling under the definition of AI systems (Article 3) and thus the scope of the regulation, the Slovenian Council Presidency has narrowed the definition accordingly. Roberto Viola, Director General of the DG Connect, had most recently defended the definition in the Commission proposal against similar criticism during the first joint debate of the lead committees in the European Parliament IMCO and LIBE (Committee on Civil Liberties, Justice and Home Affairs): It was broad enough to include the various AI systems, but also sufficiently narrow to exclude simple systems that only process data.
The French Council Presidency plans to present a new compromise proposal on the obligations of providers and users of high-risk AI systems and other stakeholders (Articles 16-29) of the regulation on February 10. Another compromise proposal on Articles 30-85 and Annexes 5-9 is currently scheduled for February 22.
The lead committees in the European Parliament, IMCO and LIBE, plan to hold a first hearing on the AI regulation with experts and scientists on March 16. Rapporteurs Brando Benifei (S&D) and Dragoş Tudorache (Renew) plan to draft their joint report by April 5.
Like after every good crisis, a sea change is currently taking place in EU economic governance. While much of the public reform discussions are focusing on fiscal rules and the Stability and Growth Pact (SGP), a more silent revolution is also going in EU economic policy coordination. In a new paper, we argue that this revolution will change EU economic governance for good. And that its current form needs a serious rethink.
It is a less known feature of the EU’s new €800 billion joint recovery fund that it has been tightly wed with the so-called European Semester. The Semester is the EU’s main tool to steer member states towards coordinating their national fiscal, labor market, or social policies. For that, the Commission monitors national economies and hands-out annual reform recommendations (CSRs) to the capitals.
Generally, the Semester has a bad reputation. It is perceived as overly bureaucratic, the process lacks sticks and carrots, member states have little incentives to follow the Commission’s advice and compliance has been low and declining. With the new Recovery and Resilience Fund (RRF), this changes.
To tap into the recovery money, capitals must submit detailed national recovery and resilience plans (NRRPs) to the Commission. These plans must not just outline how the new European money is spent but also include reform agendas that address “a substantial subset” of the CSRs issued under the Semester. The money then comes in tranches. Member states only receive the next payment if the Commission is satisfied with their investment and reform progress.
So, until the RRF ends in 2026, the Commission can now in theory hold back billions in loans and grants if member states do not follow agreed reform plans. This adds a lot of teeth to EU economic policy coordination and increases the leverage of the Commission. It will also have a lasting impact on EU economic governance. Thus, getting this right is important. We think that this requires three things.
First, the Commission should scale back recommendations under the regular Semester. By linking the fulfillment of those CSRs included in NRRPs to money, the RRF effectively makes some CSRs much more binding than others. In the next five years, all eyes will be on the national recovery and resilience plans and the implementation of the RRF. Long additional lists of CSRs outside the RRF are bound to be ignored. The Commission should therefore scale down its ambitions about what can be achieved with the Semester outside the RRF.
Second, the RRF introduces asymmetric leverages into EU economic coordination. Some member states like Hungary, Greece, Spain, or Italy are eligible to large sums of loans and grants under the RRF. These countries had to submit ambitious reform plans and will come under a lot of pressure from Brussels in the coming years.
Other countries like Germany or the Netherlands receive less and can ignore the Commission’s wishes more easily. Politically and economically, this is problematic. The Commission, thus, should not ask too much from countries in dire need of RRF funds while leaving economically-well off ones off the hook on key reforms. It should make sure that money flows only in return for progress – no matter the states or sums involved. And it should make full use of the Macroeconomic Imbalances Procedure (MIP) to keep countries that receive little RRF money and have more fiscal space like Germany on the radar.
Third, national reform and investment plans will need to be adapted. Effectively, the NRRPs lay out more or less detailed milestones and targets for each and every investment and reform that member states have committed to undertake by 2026. A lot will happen between now and 2026. Investments can fail, European priorities can evolve, and national governments can change.
If so, member states and the Commission will need to bargain whether and how to adopt plans. To turn the new Semester into an effective coordination tool, the Commission should not insist on following original plans by the letter. Instead, it should use instances of renegotiating parts of the plans as an opportunity to constantly update areas where national and European priorities overlap.
The new marriage between the RRF and the European Semester will shape EU economic governance long beyond its official expiration date in 2026. The Commission has just re-opened a public review of the EU’s economic governance framework.
The link between the RRF and the Semester means that tinkering at the margins of the old system is no longer an option. Any debate on the future of the Semester will now take place against the backdrop of the RRF experience. And any big future fiscal capacity at the EU level will almost certainly come with a Semester-style reform leg. The next years will determine what direction this discussion will take but in any case, it will change the EU’s economic landscape for good.
