AI policy careers in the EU

by Lauro Langosco11 min read11th Nov 20196 comments


AI GovernancePolicy changeEuropean Union Career choice

This article[1] is compiled from interviews and conversations I’ve had with people working in the EU, national governments, and other relevant institutions. I don't have any personal experience in the field.

My goal here is to (partly) adapt Niel Bowerman’s 80.000 Hours guide to US AI policy careers to an EU context, especially the sections ‘where to aim long term’ and ‘key routes in’. I hope this document can be useful for informing / orienting the early-stage career decisions of Europeans interested in policy careers. I'm framing everything around AI policy, but much of the content applies to careers in policy more generally.

I'll start with a short discussion of whether EU institutions are important for AI governance (for a broader discussion of this topic see Stefan Torges's post). The section ‘Key recommendations’ gives a short summary of the main takeaways, and afterwards I give some more detail on the claims and the considerations going into them.

I want to say explicitly that this isn't in any way meant to be authoritative―on the contrary, I hope this post can help encourage more discussion on these topics.

Are EU institutions a good place to work on AI policy?

(Compared to alternatives such as national government or international institutions).

Some reasons in favour

  1. Working in the EU is a great way to influence the commercial policies of the world’s second-largest trading bloc.

    Trade and competition policy are exclusive competencies of the EU. Trade policy can plausibly affect how transformative AI is developed. The effect on safety could be indirect―e.g. by increasing Europe’s competitiveness in tech―or direct, e.g. in the form of standards or regulations on safety and research transparency. Interventions can go both ways.

    Thanks to trade agreements and the member states’ wealth, EU standards are very influential internationally (for example, GDPR is effectively being adopted around the world)―see also the Brussels effect.

  2. Small member states may not play an important role.

    If you’re from a small European country then the EU might be the most powerful institution you can build a career in.

Some reasons against

  1. Working in the EU is a bet on commercial standards / trade policy.

    The EU’s competencies are somewhat limited compared to those of national governments. A career in the EU is a ‘bet’ on the efficacy of certain policy domains―commercial regulation, trade, and possibly research coordination and funding―to help to resolve problems in AI policy.

  2. It’s easier to transition from a national government to an EU institution than the other way round.

    There are multiple ‘standard’ ways to transition from national government to EU institutions, most notably entering the Council of the EU or being hired by the Commission. The other direction seems to happen more rarely.

  3. In the European Commission specifically, career advancement is tedious.

    Multiple people I’ve talked to said that career advancement in the Commission (the EU’s executive body and largest institution) is often tedious―you might end up stuck for years working in fisheries. The personal impression of multiple people I’ve talked to is that career paths in national governments tend to be more flexible. Many opportunities are unexpected, so you shouldn’t plan your career too much; a corollary of this is that flexible career paths have more upside potential.

    This is the argument I'm most uncertain about―would love to hear other's opinions here.

I think there’s a case to be made that it’s more robust―primarily because of reasons against 1) and 2) I just listed―to start a career in national government, but I’m still fairly uncertain. The rest of this document is going to focus on EU institutions without discussion of the alternatives.

Key recommendations

Where to aim long-term

Foreign policy, defense, and national AI strategies are dominated by national governments. The European Union―specifically the Commission―plays an important role in negotiating and setting commercial guidelines, standards and regulations (in negotiation with the member states). European standards are very influential internationally. The EU also does a lot of science funding.

Briefly, some positions in and around EU institutions that I think are most relevant to the development of AI are

  • European Commission

    • Head of Unit, policy officer, or advisor in directorates relevant to AI, technology, standards, and regulation.
    • Mid-level to senior positions in the Secretariat-General
  • Council of the EU (note that roles in the council are mostly staffed by civil servants from national governments)

    • Member of Coreper I or II
    • Positions in the working groups one level below Coreper
  • Parliament

    • Staffer for an MEP that’s interested in AI
    • Staffer in a relevant committee
    • Member of Parliament
  • Think tanks

  • Government affairs departments in firms working on AI or defense contractors

Key routes in

Prestigious public policy schools / the College of Europe. One person mentioned the College of Europe as a promising place to build a network. The College offers a number of competitive Master’s programs designed “to prepare [graduates] for leadership functions requiring a strategic understanding of European issues”. A large number of graduates go on to pursue a career in the EU civil service or politics.

Other public policy schools with good reputation and network in Brussels include the LSE and the Sciences Po.

Traineeships at EU institutions. Traineeships at EU institutions are generally seen as prestigious and function as a good stepping stone to a career in the EU, at a think tank, or in a public affairs department. They typically last 5 months. Some are listed here:

  • The BlueBook traineeship at the commission
  • Schuman traineeship at Parliament
  • There are traineeships at the Council and the Court of Justice as well. Here’s a whole list of traineeships all over the EU.

For more info on traineeships, see e.g. this overview article.

If you want to intern in the European Parliament you can often apply directly at an individual MEP’s office; this gives you more control over where you end up.

Apply for an AD 5 position. In the EU civil service, roles are graded AD 5-16, with entry-level roles for graduates being AD 5. Applying for an AD 5 position at the Commission is a common next step after a traineeship (see here on how to apply). Various other entry points for different profiles are detailed here.

Work in your national government to later join the Council. The Council of the EU is made up of representatives from member state governments (Ministers and civil servants).

Join a government affairs department in a firm. For entry-level positions, companies typically require 1-2 years of work experience in EU affairs, ideally in an EU institution (a 5-month traineeship is a good way to get a foot in the door here).

Join a think tank. European think tanks tend to be less well staffed and less involved in policymaking than in the US. On the other hand, a career at a think tank can be much more flexible in terms of both career advancement and the type of work you do.

Background considerations and details

Foreign affairs and security policy in Europe

European states are among the most influential countries in the world. This is due to their economic strength, their tendency to work together in blocks, well-developed relevant institutional infrastructure (for example active embassies), as well as strong presence in international organizations.

More precise statements, like comparisons to the US or China, are tricky. We could look at soft power rankings, which rank European states highly―but it’s unclear how much rankings like this reflect reality. Some hard-to assess considerations include how strongly European strategies are influenced by the US and what types of influence (e.g. military vs. economic vs. cultural) are most relevant for how transformative AI is governed.

States tend to coordinate their foreign policy in blocks. Typical coalitions are Germany-France, New Hansa, and the Visegrad 4. A tendency towards bilateralism―in particular during crises―means that national governments are much more influential than EU institutions. The EEAS does not have the tools to exert influence over member states (in contrast to the Commission, which frequently holds its own in its domains). Anecdotally: In crises, decisions tend to be taken by informal networks of individuals rather than by institutions or through formal negotiations.

Multiple people I’ve talked to that to say that it’s hard to influence foreign policy strategy, as the field is dominated by established schools of thought and practices. Especially security policy is seen as so important that there is very little room for change. The tentative prediction here is that there will likely be little regulatory framing of AI policy coming from the foreign policy establishment; their role will mostly be in the resolving of acute crises.

That said, things are hard to predict. In a field as new as AI policy there’s likely still space for framing key questions where practices haven’t been established yet. In addition, if foreign policy strategy is hard to change, that means that any change that does happen is likely to stick around for a while.

On the EU level, foreign policy is very disorganized; political tensions between member states make it very hard to reach agreement on strategic questions (consider Venezuela as an illustrative example). The exceptions here are the domains where EU institutions have remit to act, like the common trade policy.

EU Institutions

European Commission

The European Commission is the EU’s executive body and, with roughly ¾ of the EU’s staff, it’s largest institution. In the context of AI policy, the Commission’s most important role is likely that of consolidating standards and setting regulation (e.g. GDPR).

Routes in

(Not an exhaustive list by far)

  • Begin at the bottom (e.g. after a traineeship)
  • Enter as a delegate from regional government
  • Build expertise elsewhere, then apply for an advisor position

Careers in the Commission

Career progression in the Commission is generally slow. However, jobs at the commission are for life; you have enough time. One person I’ve talked to estimates that becoming the top advisor in a specific domain (e.g. Robotics and AI) will probably take less than 20 years.

Much of the concrete work of writing legislation is done at the level of Policy Officer or Head of Unit. While the broad political goals come from above, the people that draft a given piece of legislation might only be 2-3 relatively junior officers.

While it doesn’t necessarily take long to reach an influential position in the hierarchy, it seems hard to end up in the right directorate. Multiple people I talked to said that commission staffers don’t have much say in where they get promoted to―you might spend years stuck working on a random topic before moving on to where you wanted to go.

On the other hand, one person working in the commission emphasized that it’s not necessarily important which position exactly you end up in―your personal connections and reputation are a stronger determinant of which things you end up influencing.

The Commission is organized into 31 directorates-general. I’ll list some of those that seem most interesting (though I haven’t got an ‘insider view’ on this one, so this is just my outside impression):

  • The Secretariat-General

    • Coordinates the entire commission
    • (technically not a directorate but a department)
  • Communications Networks, Content and Technology

Council of the EU

In the EU legislative procedure, the Council is the EU body that represents the interests of the member state governments. It is made up of ministers and working groups from all member states.

Negotiations are held at multiple levels. The agenda for the meetings of ministers from member states is prepared by groups called Coreper I and II, who are in turn assisted by working groups. Most amendment drafting and technical discussion happens the working group level. This makes these working groups (who are made up of civil servants from member states) a promising point at which to influence policymaking.

To get into the Council working groups you’ll have to start with a civil service position in your national government. I’m not sure what the best path to follow is here (and it probably differs from country to country).

European Parliament

In the EU legislative process, the Parliament is in theory on equal footing with the Council. In practice, many people I’ve talked to say that the Council is more influential (though some disagree). I’ve heard two main narratives put forward to explain the EP’s (purported) relative lack of power. The first is that MEPs are heavily constrained by the interests of their country and other political considerations, and have little leeway to make decisions based on personal opinion. The second is that the internal divisions of the EP leave it unable to put forward a coordinated front. In any case, it seems that Parliament only rarely rejects legislation that is supported by Council and Commission.

That said, Parliament can still be a great place to work. I can see roughly three categories of promising positions in Parliament:

  • A traineeship or internship to gain early experience and build a network
  • Staff member in the office of an MEP concerned with topics relevant to AI policy, or staff member in a relevant committee
  • Elected office (MEP)

A traineeship or internship at Parliament is a great way to gain experience in Brussels and building a network. The type of work done differs between offices, so be sure to check with them first. Some internships are mostly operational, but frequently interns are also able to work on research projects or write amendments to legislation. In any case, the EP is a fast-paced environment; you’ll be taking phone calls, talking to representatives, and dealing with people blaming you for someone else’s mistakes.

Instead of (or following) an internship, you can apply for an advisory role in the office of an MEP. These roles generally fall into two categories: administration & press and legislation & policy. The latter tend to be mostly lawyers and economists who prepare and negotiate most parts of the trilogue, and propose and write amendments. Anecdotally, many of them are hired <4 years after graduating. If you can identify an MEP interested in the right policy areas this is a promising role reachable relatively early in your career. However, there don't seem to be good opportunities for further career advancement in Parliament after reaching an advisor position.

As for running for office: the expected value here depends on regional (member state) and personal factors, so I can’t really say anything about it except that it seems we should keep it in mind as an option.

Court of Justice of the EU

The Court of Justice has significant political influence. Its cases are widely read internationally and are influential in their own right due to the high quality of discussion. The court also has an important role in resolving political deadlocks.

It’s likely that in the coming years there will be a large number of cases heard that fall broadly into the domain of AI (there’s strong consumer and data protection legislation in the EU, and the current Commission plans to propose new legislation on AI). Cases create precedents that will influence how AI legislation is interpreted in the future.

As in all other institutions, the Court’s decisions on technical matters depends a lot on their in-house advisors. Becoming an advisor on AI at the Court is a potentially promising route, though I can't say much about it since I haven’t been able to talk to people at the Court so far.

One consideration against working for the Court is that having an impact there depends on a confluence of multiple factors, including (1) a case where the ruling (or precedent) is relevant for long-term AI policy, (2) legislation vague enough to leave room for uncertainty, and (3) knowledge of which ruling is favourable.

Career paths in industry: government affairs departments

I’m currently unsure about how favourably a career in lobbying compares to one in government. Here’s one potential line of reasoning: The counterfactual impact of going into a lawmaking body such as the Commission is bigger than that of a career in government affairs. This is because it’s hard to influence the company position you’re representing, while in the Commission even relatively junior staffers working on drafting legislation can have significant influence.

In the end the decision likely depends most on personal fit, on how much the company’s goals are aligned with yours, and―if they aren’t―how much you think you’ll be able to influence the company’s goals. One person I’ve spoken to estimates that after 8-10 years in a government affairs department you might have significant leeway to interpret and influence the company position.

Big tech firms are among the largest lobbying spenders in Brussels[2]. Some tech firms are inexperienced relative to firms in other industries, though this might be counteracted by heavy spending in the case of e.g. Google or Facebook.

Going into an industry other than tech to gain career capital seems like a viable option, with some caveats. Many firms in older industries have benefited from longer experience in GA. On the other hand, career capital (in particular your network) might not be transferable from one industry to another.

Defense contractors seem like another industry (one I don’t know much about) that’s interesting on AI policy grounds, as they are strong contenders for ending up at the forefront of military AI development.

Aside on management

The pyramid is weak at the top; ministers, political appointees and senior government officials typically suffer from information overload and a shallow understanding of complex policy topics (though of course this varies between individuals and positions). Multiple people I've talked to suggest that it’s a good strategy to aim “one level below”, where groups working on domain-specific legislation can to a large degree influence policy through taking care of implementation details and framing of policy issues. This seems to be true for national governments as well as the EU. This offers an opportunity for having large influence in specific domains even in relatively early career stages.

Further reading

Some open questions

  • How relevant to AI policy are positions in the European Court of Justice? Which positions in the Court are most interesting?
  • Are there domains other than trade policy (standards and regulations) and possibly research funding and coordination where we expect the EU to be able to effectively influence how transformative AI is developed?
  • What good alternatives do EU citizens have to working in the EU or in their national government (e.g. working in government in the UK or US)?

  1. Thanks to Niel Bowerman, Arsen Fazlovic, Hiski Haukkala, Sam Hilton, Nicolas Moës, Olivier Bertrand, Charlotte Siegmann, Alex Lintz, Vaidehi Agarwalla, Moritz Laurer, Stefan Torges, Thomas Metzinger, Sebastian Krier, Charlotte Stix and Philippe Thill for conversations, comments, and other valuable input. Any mistakes are mine, of course. ↩︎

  2. ↩︎


6 comments, sorted by Highlighting new comments since Today at 4:47 PM
New Comment

Thanks for writing this, it was very interesting.

Readers might be interested in the EU's AI Ethics guidelines, which various EA-type people tried (and apparently failed?) to influence in a productive direction.

A minor note:

the world’s largest trading bloc.

according to google...

  • US GDP (2018): 20.5 trillion
  • EU GDP (2018): 18.8 trillion

and presumably EU GDP, and influence on AI, will fall when the UK leaves. (If you use PPP I think China is bigger)

Can you elaborate on the EU's AI Ethics guidelines case? What did they try and why didn't they succeed?

thanks, fixed :)

Great post. Definitely an area I'd love to see further explored!

However, there don't seem to be good opportunities for further career advancement in Parliament after reaching an advisor position.

A common next step is running for election yourself, since you have the connections and know the landscape well enough to excel in debates.

As for running for office: the expected value here depends on regional (member state) and personal factors, so I can’t really say anything about it except that it seems we should keep it in mind as an option.

I think this is too quick a dismissal. Before any regional modifiers come into play, a lot can be said base value of becoming an MEP. How much power and influence can an MEP attain with regard to a single area of policy?

My intution is that an MEPs can have overwhelming influence on uncontested areas of policy, by becoming well known for their expertise within it. The first MEP that is well read and concerned with AI safety will be able to significantly influence the policies simply by being the person in power most engaged with the topic. As long as adopting your views won't generate controversy, other MEPs are more than inclined to follow your lead.

It seems to me the expected influence on AI policy disproportionally favours the first MEP to become well known within the parliament for it.

Makes sense―I agree that the base value of becoming an MEP seems really good.

In the "where to aim long-term" section, you do not mention Commissioners' cabinet of advisors. I'd be curious to know why, as I thought these roles are quite impactful (though very political). My understanding is that Commissioners' cabinet members steer policy issues, not only in terms of substance of legislative proposal but also during negotiations with the Parliament and the Council.

My guess would be that DGs & Secretariat General's freedom to influence a policy area is inversely proportional to the extent the relevant Commissioner's cabinet cares about that policy area: if the Commissioner's adviser in charge makes it her or his pet topic with clear ambitions, the DGs staff will be constrained. For example, GDPR is reportedly the brainchild of mostly one advisor to the President of the Commission. AI has become so politicized in the EU that VP Vestager & Commissioner Breton have already been quite vocal about it and their cabinet members are therefore following DGs' work on that matter very closely (of all the files in their pipeline, it is possibly the closest to a career-defining policy issue).