11 to 13 September. A weekend to research what happens once an AI system fails badly: containment, reconstructing the incident, the regulatory response and how it gets communicated. USD 2,000 in prizes, and applications close on 6 September.
We are the community working on it from Colombia: training, research and open events you can walk into today.
This is not a discussion about robots or about a distant future. It is the work of getting increasingly capable systems to do what is expected of them, of making sure whoever uses them understands their limits, and of having rules in place before they cause harm. It runs from technical research through to public policy.
Read the long versionEvery year models do things their own builders did not anticipate. Understanding what they compute inside, getting them to pursue what we want and keeping them under supervision are still open research problems.
Hiring, credit, health, public services. Those systems already run in Colombia, almost always bought abroad and tuned on data that is not ours.
Much of AI governance will be settled over the next few years, and very few people from the region are at that table. Training the people who could sit there is part of the work, not an annex.
Training and group work for people who want to get into this seriously, without needing to have started earlier.
See everything that is open
Open talks, workshops and discussion dinners in Bogotá.
Meetings with researchers and with policy people already working on the subject in the country.
Organisations we sustain continuous work with. We have also received funding from all four. The partners of each hackathon are announced in its own call.
Convenes AI safety research hackathons, open to participants anywhere in the world.
Their free courses are the standard way into the field. Through Rapid Grants they fund concrete work: USD 1.4 million awarded in total, with decisions in three days on average.
Mentorship programmes and support for university AI safety groups.
In 2024 alone it committed around USD 50 million to technical AI safety research. Until 2025 it was called Open Philanthropy.You do not need to be a researcher, and you do not need to have studied artificial intelligence.
