The convergence of AI and Free Software is briefly analysed in this paper from the perspective of licensing, especially considering the necessity to promote openness respecting the historical principles of software freedom. Some challenges in achieving openness in the licensing scheme of AI are highlighted. Inparticular, the uncoordinated growing proliferation of licences claiming to be “free and open source” but imposing extra limitations on software freedom may lead to licence incompatibility, a well-known challenge faced by the Free Software community.This paper clarifies why using licences to address behavioural and usage restrictions may affect distribution of control over AI technologies, and how the complexities of non-free licences in multi-source software development can affect compliance efforts.In conclusion, three recommendations are proposed:• Preserving openness in AI by safeguarding the four freedoms of software. Restrictions on software freedom disable control, transparency and oversight over technology. This results in a negative impact on people’s digital autonomy, distribution of powerin the society and ultimately the democracy.• Keeping licensing of AI technologies cohesive and interoperable with Free Software licences by avoiding licence proliferation, increasing legal interoperability and simplifying licence adoption.• Encouraging engagement with civil society actors in initiatives aimed to make AI more open, accessible, transparent and auditable.