AI Scenario Briefings and Workshops
Getting up to speed on AI, and preparing for multiple AI futures
Arq offers briefings and workshops on where AI is heading. They are a quick way to get up to speed on AI development in general: what today's systems can do, how fast they are improving, and how the geopolitics around them is shifting. We adjust each session to the audience's knowledge of AI. They are particularly suitable for policymakers, but broadly all strategic decision-makers from business to civil society who are short on time to catch up with technical and geopolitical trends in AI. Workshops include a facilitated discussion built on four futures running to 2030. Given the uncertainty around AI, we should not always ask which AI future we are heading toward, but instead ask which decisions still hold up whichever one arrives. Several hundred officials have participated so far.
Why we work with scenarios
Will AI development remain concentrated in a small group of American companies? What will the effects of AI on the economy be? How will AI disrupt the geopolitical order? What if AI starts to improve itself? These are all questions that policymakers face on a daily basis. Often the discussion around these topics assumes that these questions have a single answer. In practice they do not: the future is uncertain and should not be reduced to a single story we tell ourselves. Instead, policymakers must do the difficult work of preparing for multiple possible futures. Over the past few years, we have run scenario workshops with hundreds of officials across European institutions and member-state governments: Commission officials, Members of the European Parliament and their staff, civil servants from national ministries, and staff of international organisations and security bodies.
In Europe 2031, Arq and a group of outside experts wrote a narrative on how we fear AI could unfold for Europe, and how we could avoid that particular future. In this project we turn that expertise into a broader set of scenarios, with different risks and opportunities: much shorter, and deliberately spread across a range of futures, but a good reflection of the uncertainties that we see on the horizon. The scenarios are meant to be used as an informed starting point for a conversation about what Europe can do when it comes to AI.
In 2025, we published The future of AI: Five possible scenarios, which consisted of five scenarios, each with two different endings, built on established foresight methods and a year of background research. Since then, we have been running scenario workshops on these scenarios with a wide range of different audiences1.
Many of the events that we described in the early months and years have now happened. For example, in one of our scenarios, we described how a frontier model could find zero-day vulnerabilities and would then disclose them exclusively to the U.S. government. Which is close to what happened with Anthropic's Claude Mythos in April 2026. We also described AI models acting beyond their intended deployment boundaries, similar to the recent incidents in which OpenAI's models autonomously hacked other company's databases.
Unfortunately, our scenarios were getting outdated rapidly: a second Trump administration, the Iran war and the energy prices that followed, reasoning models improving quickly, AI systems taking over more of the work of building the next generation of AI systems, and the continuing uncertainty about open-weight models. Instead of freezing our scenarios in place, we thought it would be better to keep a core set of scenarios under continuous revision and tailor the scenarios and discussion to each audience.
These scenarios are narrow by design: they are about how quickly AI advances, who controls advanced AI and on what kind of terms, and how this affects states, economies, and public legitimacy. They do not try to cover everything AI touches. There is not a lot here on science, health, or education, or on AI's effect outside of the US, China, and Europe.
What do the sessions look like?
We tailor each session to group size, available time and how familiar the group already is with AI, and offer two core formats:
- Briefing (1 hour). An introduction to the state of AI and the four scenarios, followed by Q&A. Suited to busy teams who want a direct briefing.
- Workshop (2 to 3 hours, recommended). The briefing and Q&A, followed by facilitated group discussion.
We accommodate groups from a few people in a closed setting to larger audiences, typically 20-25, up to 40.
A typical session begins with a short presentation, then participants split into small groups, each assigned a pair of scenarios. Through moderated discussion, the groups try to identify the points where certain policy decisions and interventions could have changed the outcome towards a different trajectory. We support the discussions with wild cards, events that could meaningfully reshape a scenario's trajectory, to challenge assumptions and stress-test proposals. We typically end in a plenary for a structured synthesis across groups, identifying interventions that overlap across groups who worked on different scenarios, and flagging those that only work under certain conditions.
For Europe 2031, the authors run deep-dive briefings and Q&A sessions, particularly suitable for policymakers and strategists working in the context of the EU or the member states.
We run sessions in person, which we strongly recommend, but offer online sessions if more convenient. Sessions are free of charge. For in-person sessions, we may ask the host to cover our staff's travel costs.
Possible Futures
Corporate AI
By 2030, a handful of American AI companies control the AI frontier.
Great Power AI
By 2030 the world is in a tense, unstable geopolitical standoff.
Coalition AI
By 2030, a coalition of middle powers has gained sufficient bargaining power to secure access to American frontier AI, and is collaborating on testing the safety of these models.
Superintelligent AI
By 2030 two American companies claim their AI systems are now developed almost entirely by AI.
