OSS Week 2026: AI Between Hype and Stance – Part 1

In mid-June, we held our Open Source Software (OSS) Week – our annual internal “tinkering week” for open source, professional development, and a shared focus on an exciting topic.

This year, Artificial Intelligence (AI) was at the center, with a focus on how we can benefit from these tools in our daily work while remaining true to our convictions. We therefore aimed to examine this technical innovation through the lens of the social and material contexts in which it arises.

In this first part of this two-part series, we look back at the ethical, ecological, and philosophical discussions that shaped our week. In the second part will cover the concrete technical projects and hands-on learnings.

Infrastructure: A Compromise with a Sense of Proportion

In thelead up to the OSS week, the question of which infrastructure to use for our experiments occupied us. On ethical grounds, we ruled out providers like OpenAI and Google, but we still required an industry-standard, closed-source benchmark. To this end, we included Claude from Anthropic in our tests – not as an ethical benchmark, but as a deliberately chosen reference against which we could evaluate open-weight models. For it is these open models from which we expect greater independence and digital self-determination.

Local hosting was ruled out for ecological and economic reasons: The acquisition costs for adequate hardware is unfeasibly expensive, the lifecycle of the components was unclear, and the risk of rapid technical obsolescence is high.

As a pragmatic and values-consistent compromise, we therefore chose cloud hosting the open-weight models in a solar-powered data center run by Infomaniak¹ in Geneva. This allows us to avoid direct dependency on international mega-corporations, remain flexible in our setup, and reduce the environmental impact of our experiments through the use of renewable energy.

Stance Over Hype: Our Conversations on Responsibility and Sovereignty

The choice of our infrastructure, however, was only the starting point for us. Whether at the morning stand-up, over a shared meal, or later in the evening – many moments of exchange arose in which we challenged each other and encountered a wide variety of perspectives.

We are happy to share a few insights from these shared conversations with you below.

Transparency as Non-Negotiable

How do we deal with AI-generated code, and who bears the responsibility?

Our current stance on this is: The use of external service providers like Anthropic must be communicated transparently to customers, partners,and contributors to our open-source projects. We regard the use of AI on digital infrastructure under our control as equivalent to using any other internal tool and not as outsourcing.

Between the Promise of Productivity and Responsibility for Our Environment

Does the efficiency gain from AI-generated code stand in proportion to its ecological and social footprint?

Our conclusion was sobering: in general, the total costs – CO₂ emissions, water consumption, land use, and not least the working conditions in training the models in the Global South – are not commensurate with the productivity saved².

However, by deliberately choosing regional, solar-powered infrastructure in Geneva, we make a significant contribution to resource-conserving use. What remains even with a relatively sustainable setup are the material and ethical costs of manufacturing the hardware, as well as data acquisition and the training of the models themselves. we will not be able to eliminate these fundamental problems of the industry within our framework for action, but we can minimize their impact in relation to our own conduct.

Dependency and Economic Sovereignty

Are we making ourselves dependent on proprietary AI providers and their pricing policies in the long run?

Thanks to our long-standing partnership with Infomaniak, we have direct access to the hardware and comprehensible billing per token. This protects us from opaque pricing models and subtle performance reductions. Moreover, we consider the hosting in Geneva an ideal compromise: we retain autonomy without making expensive capital investments in rapidly changing hardware, and instead benefit from a professional host that operates its infrastructure continuously and can utilize it over its entire lifetime.

Where Do We Go From Here?

Our OSS Week 2026 has shown that careful engagement with AI is less a technical challenge than a question of stance. We have not found a perfect solution, but a clear draft for our future course of action. We remain committed to transparent processes, rely on climate-compatible infrastructure, and use powerful open-weight models where it makes sense.

In the next part of this series, we will then take a look at some of the technical learnings we discussed during that week.

We look forward to staying in dialogue with you. What questions occupy you in your dealings with AI? We warmly invite you to share your perspectives with us.

1: https://www.infomaniak.com/de/hosting/ai-services/open-source-models

2: UNU-INWEH Report: Aczel, M., Chamanara, S., Matin, M., Farsi, A., Marwala, T., Madani, K. (2026). Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints. United Nations University Institute for Water, Environment and Health (UNU-INWEH), Richmond Hill, Ontario, Canada. doi: 10.53328/INR26RMA002