As AI tools become increasingly prevalent in our digital world, we are only gradually becoming aware of the wide-ranging consequences of the ‘AI revolution’. Most people will by now have realised that the provision of generative AI comes with high energy and water consumption. However, like many modern platforms, ChatGPT, Gemini and the like are also designed to keep users engaged. They prompt users to ask more and more questions and taking on more and more tasks.
This product design is based, amongst other things, on dark patterns. This is a user interface that has been specifically designed to trick users into taking certain actions. It also uses subscription models, makes it difficult to use generative AI in an environmentally responsible way. The new aim2balance.ai platform, on the other hand, uses a prepaid model and instructs the underlying language models to ask fewer follow-up questions.
Could platforms like this mean that AI finally stays within planetary boundaries?
What exactly is aim2balance.ai?
“A few years ago [in the Bergwaldprojekt], we realised that a new technology was on the horizon that could really help us in our work,” says Stephen Wehner as he begins our conversation a few weeks after the official launch of aim2balance.ai. The Bergwaldprojekt (Mountain Forest Project) is an environmental protection organisation that works on forest ecosystems.
“In the forest,” Stephen continues, “AI systems are increasingly being used. And the sooner we can identify problems such as wildfires or drought, the sooner we can respond. But when we tried out the first applications, we quickly realised it has loads of drawbacks!”
Stephen Wehner and his team at wanted to create an AI tool that would eliminate many of these drawbacks. Their approach stems from ecology itself, explains Wehner:
“In nature, diversity always leads to greater resilience – that’s something we’ve learnt in 30 years of running the organisation. The business world, however, always focuses on efficiency, and when a problem arises, it is usually exacerbated. As a not-for-profit organisation that doesn’t have to optimise for efficiency, we can therefore offer an alternative such as aim2balance.ai.”
Thanks to press access, we were able to try out the platform for ourselves.
The platform essentially works in the same way as established solutions. When you open aim2balance.ai in your browser or in the app, you’ll find the familiar chat window with plenty of space for a conversation history. Experienced AI users will be familiar with personalised assistants and the ability to tailor responses to their own specific needs. Like any modern ‘AI’, Aim2Balance can generate not only text content but also images and code, as the responses are based on established large language models.
“Every query is first processed by a very small AI model,” replies Stephen Wehner when asked which language models aim2balance.ai uses. “This model selects which of the integrated models can handle the query most efficiently. This alone enables us to reduce our resource consumption by at least 50 percent.”
Why use AI heavyweights for a green AI tool?
The resource consumption of large language models has many dimensions. In the early days of the technology, criticism regarding a lack of sustainability mostly focused on resource-intensive training. LLMs are still being trained using trillions of parameters in huge data centres. However, studies have shown that the usage phase accounts for the larger share of a language model’s environmental footprint over its life cycle. The so-called inference phase—that is, everyday use—is therefore also the key lever in the sustainable use of AI. This is where aim2balance.ai comes in.
As Stephen Wehner explains, his team employs several strategies to reduce resource consumption. “As we are hosted in Europe, we are already operating in line with the GDPR as a matter of course. We also transmit all data using encryption codes that are also used for banking data. This ensures that no data is passed on to big tech companies.” The aim2balance.ai team can therefore avoid analysing usage data. Miriam Ruhenstroth from the Institute for Technology and Journalism revealed to us in another interview that this can be a real power guzzler.
The servers mentioned are also powered entirely by green electricity, and aim2balance.ai channels 2.5 percent of its revenue into nature restoration projects. According to Stephen Wehner, there is a simple reason why this rather curious figure was chosen. “We asked ourselves: which indicator best describes how resources on our planet are becoming scarcer? From an economic perspective, that is the inflation rate – which is why we always use the long-term global inflation rate. And that happens to be 2.5 percent.”
The Bergwaldprojekt relies on modern language models such as Google’s Gemma 4, Mistral, Llama, Qwen, GLM 5 and Kimi K2. Although their use in a green AI initiative may seem contradictory at first glance, Stephen Wehner believes it is still necessary.
“The far more satisfactory solution would be to have European models designed according to ‘good’ principles and whose economic interests align with our principles.” However, it is precisely these solutions that are lacking. As the association does not have the necessary resources to train its own language model, it is focusing on inference for the time being.
aim2balance.ai takes psychological factors into account
One of aim2balance.ai’s key features is the way the platform encourages users to adopt a more mindful approach to using AI tools. The enormous proliferation of AI tools since 2023 continues to be driven mainly by big tech companies in the US. Although there are an increasing number of alternatives, such as the Apertus language model which was developed with an ethical approach and using green electricity, their uptake is not comparable to that of large models from the US or China.
And their business models are partly based on psychological and psychosocial principles designed to increase usage. So-called ‘dark patterns’, for example, result in users becoming more heavily tied to certain platforms. With the extensive personalisation offered by AI software, this dependence is increasing, as people form parasocial relationships with services such as ChatGPT. On the one hand, this can lead to an accumulation of negative psychological consequences for users. On the other hand, it also leads to more intensive use and, consequently, greater consumption of resources.
“Part of aim2balance.ai is also a level of reflection [… on one’s own use of AI.] Other chatbots always encourage users to prolong the conversation. As people are naturally curious, we’re trying to break that pattern.” Stephen Wehner goes on to explain that there is therefore an additional option in a sidebar that allows users to reflect on their own behaviour. Using the EcoLogits software, aim2balance.ai also compares energy and water consumption with that of conventional AI chatbots.
How much does aim2balance.ai actually cost?
According to aim2balance.ai’s pricing guide, €5 is enough for 450 prompts, i.e. queries to the chatbot.
Based on an average usage of 15 prompts per day over the course of a month, €5 is sufficient to cover the monthly cost. Anyone who needs more can also enter €10, €20 or a custom amount.
An automatic top-up option is also available – however, unlike with a subscription model, the topped-up credit does not expire.
aim2balance.ai also deliberately asks fewer follow-up questions in conversations and starts new accounts with a survey to personalise the AI’s output. For example, we were able to disable the use of emojis and ask the software to provide answers that are as brief as possible, without ‘polite phrases’.
Unlike established alternatives, aim2balance.ai offers neither a free service nor a subscription model. Stephen Wehner explains why this is intentional as follows: “It simply isn’t true that free services are really free. If I want to use something, I have to take responsibility for it and pay for it.” To give users more control over their spending, aim2balance.ai operates via a prepaid account whose balance does not expire: “With us, you only pay for what you’ve actually used.”
What’s next for aim2balance.ai?
If you’d like to try aim2balance.ai straight away, you can do so directly in your browser or via the aim2balance.ai app for iOS and Android. With 25 free prompts, you can explore the platform and then top up your account. There are no restrictions on the range of features available.
Compared to other platforms, it also lacks convenience features such as a live chat for conversations with the AI assistant or the ability to generate video content. Stephen Wehner comments: “aim2balance.ai is by no means intended to be the next ChatGPT. And, of course, the platform isn’t perfect yet. But the more people use it, the better it will become.”
The Bergwaldprojekt also aims to future-proof aim2balance.ai by making it open-source. That way, even if not enough people use the platform and the organisation can no longer offer it, it can continue to exist.
We would like to thank Stephen Wehner for the interview.
