In today’s edition, a rotation between your storytellers will give you a different perspective on the NATO summit in Vilnius. Also, we will try to analyse the social impact of the EU AI Act. Finally, in today’s Diplo Focus, we bring you into the campaign of a Republican outsider. Enjoy our analysis!

The Alliance of the procrastination

The NATO summit in Vilnius ended a few days ago. Ukraine received much new military aid. And – again – a promise of admission in the Atlantic Alliance. Volodymyr Zelensky showed signs of impatience with Western ambiguity: leaders of the major NATO countries have opposed Kyiv’s demands with an attitude cloaked in apparent pragmatism. But if, despite everything, a sour taste remains today, it is not only because NATO/Ukraine communication errors have provided the Kremlin’s narrative with a delicious assist, a slam dunk right there.

Was President Zelensky wrong to bet his chips on Ukraine joining NATO with the war on? This much is clear. Did NATO countries make the mistake of underestimation by allowing Vilnius to be expected for months as the watershed summit for Ukraine’s future in the Alliance? This is equally obvious. Yet what appears to be a significant weakness today is the compromise between the heads of state and government meeting for two days in Lithuania. It was an excess of caution. And perhaps, what is more serious, a strategic error.

Thus, resound in the mind of your storyteller a few words, those with which Boris Johnson opened his latest editorial for the Daily Mail: so what were they frightened of?

Inside the European AI Act: de-risking against the time

The European Union is trying to regulate artificial intelligence for the first time. This is a complex but old exercise: trying to regulate disruptive innovation to the benefit of the people is, probably, the fundamental challenge of capitalism since the beginning of the time. Complicating the EU’s attempt, however, is also the scope of a regulation that, to be effective, must be recognised globally. Otherwise, it risks being easily circumvented.

After the European Commission revealed its long-awaited proposal for regulating artificial intelligence in April 2021, there has been significant progress made by EU institutions and lawmakers in developing the EU Artificial Intelligence Act (AI Act). This Act aims to establish a comprehensive legislative framework for AI and address its ethical and human implications. Once implemented, it will be mandatory for all 27 EU Member States to adhere to its provisions.

The panorama of companies currently developing disruptive AI goes from the United States to China. And the United States, for the time being, has no intention of putting the brakes on research and development: as much as the Biden administration is aware of the risks, after so much talk, all that has been done to date – in fact – is a series of non-legislative measures suggesting to the AI giants how they should behave to self-regulate their technology.

The idea of the EU is to classify not so much the algorithms as the intensity of the risk related to their use: low, medium, high, and unacceptable. It is a very useful first step to ban those technologies that have too high a risk in terms of violating people’s fundamental rights.

If, for instance, AI blocks spam, the risk is low. If, on the other hand, it gives you a medical diagnosis, the risk is clearly much higher. Not to mention the possibility of AI being used to do social scoring, i.e., to rank people’s reputations (as China does for its citizens) – and the risk for this use would be classified as ‘unacceptable’.

However, there is at least one problem with this rule. Since AI develops rapidly, it is not always easy to tell in advance whether a particular type of software is dangerous or not. One example above all: the case of chatbots. Whereas until a couple of years ago, they would have been classified as low-risk algorithms, today, with Chat-GPT, the scenario has radically changed.

It’s important to note that these rules are still in the lengthy EU legislative process and might take at least two years to enforce fully; they were drafted before the latest surge in generative AI technology, which can potentially create more challenges for policymakers. As a result, Brussels has a lot of work to do to catch up with the rapid advancements in AI and address the many potential implications.

On the other hand, we need to acknowledge that there will never be a perfect law, and any regulation passed will have to be – in some way – flexible. Able to adapt, or be adapted, to changes in society and technological developments.

Perhaps even more critical than having flexible regulation is the urgency of having shared standards. In fact, as much as the AI Act can cover regulation in the EU, it cannot do so for AI developed in other nations, posing enormous problems regarding the dynamic of importing and exporting technology in a hyperconnected world.

The proposal to reach global cooperation is not new. There have been early discussions in the G7 and plans to include receptive countries like India, Indonesia and Brazil to move towards a global agreement on AI regulation and security protocols, but not fast enough. These are only small steps, at least for now.

The issue is broad and complex, of course, but there are many possible solutions to ensure that there is ethical control of artificial intelligence: one of the proposals in the draft AI Act is, for instance, to create processes that make use of synthetic data.

Synthetic data is built from real data and generates new data that retains the statistical properties of the original data, so on the one hand it allows to anonymise the results (and thus not violate privacy), and on the other hand to act on bias reduction. This is also why there is a lot of focus in the AI Act on the data underlying artificial intelligence, its quality, and its validation in terms of robustness and representation, as well as the requirements of data privacy and data minimisation. Synthetic data have the potential to play a key role in the realisation of these requirements towards secure and reliable AI.

The European AI Act can be the forerunner in establishing comprehensive and enforceable rules to ensure AI security and responsible innovation. Despite some limitations and although it might take 2 to 3 years to come into force, there are nevertheless several reasons to believe that it has the potential to create a benchmark that can be useful for cooperation in the global regulation of AI.

One thing is sure: the time window for the regulation of AI is rapidly shrinking. There will be no nuclear explosion, but it is like a snowball rolling downhill that is getting bigger and bigger and harder and harder to stop. That is why it is important to act now, as Europe is already doing, but on a global level. At least, when it comes to respectable democracies.

Diplo Focus: The candidate you don’t expect. Mike Pence, inside the political life of the Republican Outsider

On January 6, 2021, hiding for interminable minutes in the basement of the Capitol, trying to evade the angry mob that wanted his head, one wonders if Mike Pence finally thought of his father. Edward Pence, a man so stern that he has never forgiven himself for outliving his comrades-in-arms in the Korean War.

One thing for sure, Edward Pence must have taught the young Mike: what makes you the person you are is how you will behave when the wind blows hard and against you. It will be there, at that exact moment, that you will have to prove your value.

‘I am a Christian, a conservative, a Republican: in that order,’ he likes to repeat. There is an anecdote: Mike Pence became a Republican the day after voting for a Democratic president, Jimmy Carter. The year is 1980. His new history teacher, Professor Curtis, shows him that Ronald Reagan is not the ‘vacuous movie star’ Pence thought he was.

This lesson may come in handy today: nothing is more complicated than shaking a label off. Let alone if that label says ‘traitor’ on it. Meeting thousands of voters in his long run for the Republican nomination, there is no day when someone does not accuse him of betraying Donald Trump and his people. And of being the only one truly responsible for Joe Biden being in the White House.

To one of these voters in Iowa, he responded with an unexpected flicker of pride: “Thank God, I know I did exactly what the Constitution of the United States asked me to do that day. I have kept my oath.”

Polls show him in third place nationally, too many points behind Donald Trump and his first challenger, Ron DeSantis.

A few days ago, hosted by Republican anchor Tucker Carlson, back from a surprise trip to Ukraine, Mike Pence refused to be roasted on a spit.

Some would cry out for a miracle if he won the nomination and the presidency. On the large fireplace of the Vice-President’s residence in Washington, Mike Pence had a plaque engraved with a passage from the Bible: ‘For I know the plans I have for you,’ says the Lord, ‘plans to prosper you and not to harm you, plans to give you hope and a future.’

In life, sometimes you just must believe.