Berlin AI Safety

Berlin AI Safety (aisafety.berlin) is a Berlin-based AI-safety community group whose central activity is a monthly public meetup bringing together researchers, policymakers, and technically-minded practitioners in Germany around the AI-safety literature and related governance questions. Founded around 2022 by Manuel Allgaier — former director of Effective Altruism Germany and organiser of EAGxBerlin 2022 — alongside co-organisers including Guy P (handle: GuyP / broccoli.boo) and Alex McKenzie, the group operates as the insider-community complement to PauseAI Germany's public-protest work: where PauseAI Deutschland mobilises public protest around a treaty-backed AI pause demand anchored in Halle (Saale), Berlin AI Safety cultivates the research and policy practitioner community within which a serious German AI-safety conversation can develop. It is the corpus's first insider AI-safety community group entry in continental Europe.

Founding and structure

The group's earliest documented EA Forum event is a Berlin AI Safety Open Meetup posted for 30 July 2022, establishing that month as the practical founding anchor. Manuel Allgaier, the co-founder named on the group's website, served as director of Effective Altruism Germany and organised EAGxBerlin 2022 — a 900-person EA conference in Berlin — before helping launch Berlin AI Safety; his trajectory from EA-community organising to specialist AI-safety meetup mirrors a pattern visible across the EA-to-AI-safety community-building layer in other European cities. Alex McKenzie, listed as an organiser on the aisafety.berlin website, is a Berlin-based AI alignment researcher formerly at LASR Labs (a safety-skilling programme) and now leading AI alignment research at AE Studio. Guy P posts the majority of the group's event announcements on both the EA Forum and LessWrong under the handle GuyP, identifying himself on the EA Forum as a Berlin-based AI researcher and aisafety.berlin organiser.

Community infrastructure is extensive for a volunteer-run group: a Luma events calendar, Telegram announcement and community-chat channels, a Signal group, and WhatsApp — all linked from the main website. The group has no announced parent organisation.

Monthly meetups

The group's primary recurring format is a monthly open meetup — a Tuesday or Wednesday evening event running from approximately 7:30 pm (doors) through 10:30 pm (open networking), with a structured core: a researcher talk or paper discussion (7:45 pm), followed by a community announcements and pitch round (8:30 pm). Recurring venues include Thoughtworks in Friedrichshain and the Teamwork Space on Müllerstraße. AISafety.com rates the group "Very Active".

The meetup programme covers the breadth of the AI-safety literature, from technical alignment to governance and policy. Documented sessions from 2024–2025 include: a November 2024 talk on scalable oversight by Yoav Tzfati, a MATS (ML Alignment Theory Scholars) alumnus; a March 2025 discussion of the Emergent Misalignment paper; an April 2025 session on "What Failure Looks Like" (AI risk scenarios); and a July 2025 talk on securing frontier AI models against powerful adversaries, drawn from SL5 Task Force and RAND Corporation research. Event listings are cross-posted to both the EA Forum and LessWrong, reflecting the group's dual positioning inside EA-adjacent and rationalist / LessWrong AI-safety community networks.

AI Futures Forum and broader events

Beyond the monthly meetup, the group co-organised the AI Futures Forum — a day-conference held on 21 June 2025 in collaboration with BlueDot Impact, framed as an opportunity for participants to "connect with Berlin's AI Safety & Governance ecosystem" and built around the theme of Germany as an "AI Middle Power" country. The event reflects a strategic framing: positioning Germany — with its AI industry, regulatory infrastructure, and proximity to EU AI policy — as a key battleground for AI-governance questions, distinct from both the Anglophone frontier-lab debate and the Brussels-centred EU regulatory process.

The group also co-hosts periodic AI Safety Coworking sessions with the Æthos Foundation and the Foresight Institute, and has organised occasional hackathons and retreats supplementing the monthly meetup cadence.

Place in the movement

Berlin AI Safety is the corpus's first continental-European entry for an insider AI-safety community group — one whose central activity is expert talks, paper reading, and practitioner networking rather than public-facing campaign or protest organising. It sits in the Germany coverage alongside PauseAI Germany but serves a structurally distinct function: where PauseAI Germany runs a public Bürgerbewegung (citizens' movement) targeting the German government's AI summit positions with academic petition signatures and street protest, Berlin AI Safety cultivates the practitioner layer — the Berlin-based researchers, policymakers, and engineers who track the technical safety and governance literature — through regular low-barrier access to expert talks and peer-discussion forums. The two groups are not the same audience; they are complementary layers of the same broader German AI-safety ecosystem.

The insider-community pattern the group represents — a monthly expert meetup routing practitioners from EA, LessWrong, and adjacent networks into the AI-safety literature — parallels the formation of AI Safety 東京 in Japan and AI Safety South Africa in Cape Town: all three groups share the benkyoukai-style "accessible technical reading group as community on-ramp" structure, adapted to their respective local practitioner communities. Berlin's specific contribution to the federation's AI-safety organising layer is its positioning at the intersection of Germany's engineering and research ecosystem and EU-level AI-governance politics — a combination that the AI Futures Forum's "Germany as AI Middle Power" framing reflects explicitly.