Making AI safe: What you can actually do
Anyone who has followed the news about artificial intelligence over the past few months could easily come away uneasy. Media around the world have reported extensively on this year’s worrying AI incidents. The systems are becoming more capable at a rapid pace. At the same time, tests and incidents keep showing that we neither fully understand them nor can reliably control them. Billions flow into building ever more powerful models, and only a fraction of that into making them safe.
Still, we are not powerless. A small but highly qualified community of researchers and policy experts is working to keep advanced AI under control. Because the field is still underfunded relative to its importance, additional money can achieve a great deal here. If you want to donate, you can do so through Effektiv Spenden, tax-deductible in Switzerland and in Germany, even when the organisations we fund are based abroad. Our donation tax calculator shows how much you save.
Our giving fund «Ensuring Safe AI»
The way to do this is our giving fund «Ensuring Safe AI». It has a single goal: to reduce the risks from advanced AI that could have catastrophic consequences. We pool the donations we receive and pass them on twice a year to the organisations where additional funding does the most good right now.
Through our giving funds, we have already passed on more than 1.2 million Swiss francs since 2023 to organisations working on AI safety. The money comes from donors in Switzerland, Germany and Austria.
Why we rely on Longview Philanthropy
We do not decide ourselves where the money from the fund goes. We follow only the recommendations of our partner Longview Philanthropy. There are good reasons for this.
Longview is a nonprofit philanthropy advisor that specialises in reducing global catastrophic risks. Its AI team is among the largest and most experienced grantmaking teams in this field. It has six staff and has been led by Michael Aird since January 2026. The team follows the field continuously, examines organisations in detail and is in close contact with researchers, policymakers and other funders. This lets it see where an organisation can put additional money to real use and where enough funding is already in place.
When you donate to the fund, you benefit from the same expertise that is otherwise reserved for large foundations and philanthropists.
The organisations we have funded so far
METR
- Grant: around CHF 123,000 (H1/2023)
METR (Model Evaluation and Threat Research), based in Berkeley, grew out of the evaluations team of the Alignment Research Center and has been independent since late 2023. It was founded by former OpenAI researcher Beth Barnes. METR did much to shape so-called «evals»: systematic tests of which dangerous capabilities an AI system has. METR has tested models from OpenAI, Anthropic, Google DeepMind and others before their release. A 2025 study drew wide attention. It shows that the length of tasks AI systems can complete on their own doubles roughly every seven months. Newer data even point to a faster pace. To stay independent, METR accepts no money from AI companies. It also provides technical support to the EU AI Office.
RAND – Center on Technology and Security Policy (TASP)
- Grant: around CHF 146,000 (H2/2023)
RAND was founded in 1948 and, with around 2,000 staff, is one of the most influential think tanks in the world. The TASP center studies how dual-use technologies such as AI and biotechnology are changing international security. From this it develops concrete policy options for governments in the US, in Europe and among their partners, for example on controlling AI chips and computing power. Because RAND works directly with policymakers, this research can feed straight into policy.
CeSIA – Centre pour la Sécurité de l’IA
- Grant: around CHF 183,000 (H1/2024)
The French Center for AI Safety combines public education, training and policy work. CeSIA initiated the «Global Call for AI Red Lines», which was presented at the United Nations in September 2025. It calls for internationally binding red lines that AI systems must not cross. CeSIA also informs the French public about the risks and opportunities of developing AI safely.
The Future Society
- Grant: around CHF 109,000 (H2/2024)
The Future Society works on the regulation of advanced AI models. It played a major part in drafting the EU Code of Practice for general-purpose AI models, a central instrument for implementing the EU AI Act.
SaferAI
- Grant: around CHF 160,000 (H1/2025)
SaferAI is an independent research institute based in Paris. It applies proven risk management methods from high-risk industries such as aviation and finance to AI. In a public rating, SaferAI assesses how well the leading AI companies manage their risks. It has also published the first independent European risk evaluation of a frontier model. SaferAI’s influence on regulation is especially large: in the European standards body CEN-CENELEC, it leads the work on the risk management standard for high-risk AI. It took part in all four working groups on the EU Code of Practice and was the only AI safety organisation appointed to an OECD task force.
Safe AI Forum (SAIF)
- Grant: around CHF 326,000 (H2/2025)
SAIF promotes international scientific cooperation on AI safety. At its core are the «International Dialogues on AI Safety» (IDAIS), where leading AI researchers from the US, China, the UK and the EU work out shared positions on the greatest risks. At a time of growing geopolitical tension, this is one of the few reliable channels of dialogue between the most important AI nations.
The grants from the first half of 2026 are not yet listed here. After the end of each half-year, we first reconcile all incoming donations. On that basis, Longview prepares its recommendations. This is followed by a legal review and coordination with the organisations. We will publish the grants from the first half of 2026 in two to three weeks.
Close to the community, in Switzerland too
How lively the AI safety community has become in this country was shown by the Swiss AI Safety Days on 7 and 8 November in the main building of ETH Zurich. The event is an official part of the «Road to Geneva» leading up to the 2027 AI summit in Geneva. The highlight was the keynote by Yoshua Bengio, Turing Award winner, professor at the Université de Montréal and the most cited AI researcher in the world. Other speakers included Stuart Russell (UC Berkeley) and Adam Gleave (FAR.AI), as well as experts from Google DeepMind, Anthropic, ETH and EPFL. The first edition the year before had already been named the best of more than 230 events at Swiss AI Weeks 2025.
Through events like these and our partnership with Longview, we are closely connected to the community. If you have questions about a particular organisation or approach, we are happy to help or put you in touch.
Personal donation advice
Would you like to learn more, are you thinking about a larger donation, or are you unsure which way is right for you? Then sign up for a free donation consultation. We are glad to take the time for you.
About the author
Country Manager Switzerland
More blog posts on AI
Get our newsletter
We share tips and news on how you can improve the world effectively.
15,779 people already read it.
The form could not load, probably because of an ad or script blocker. Turn it off for this page or write to us at info@effektiv-spenden.org.