At BOSC 2026, we want to talk about the elephant in the open-source room: Is generative AI an advantage or a hindrance to open source?
We invite abstracts on this topic. Some might be selected to give talks at BOSC (which will be part of ISMB 2026). We may also invite some of the chosen speakers to participate in a panel. The submission deadline is April 9.
For example, here are some possible topics (but don’t feel restricted to these):
- Reuse: how can we encourage and facilitate reuse of tools and frameworks when AI makes it easy to code things up from scratch?
- Evaluating open source projects: AI tools can generate thousands of lines of code in seconds. The most costly process is now verifying that code for scientific accuracy (https://arxiv.org/abs/2507.09089). What are some good approaches to address this?
- Contribution guidelines: balancing scale and utility of AI-assisted development with community-building
- How should an open source project assess pull requests from AI agents?
- Are zero-tolerance bans on submissions generated using AI reasonable? (e.g., https://medium.com/@livewyer/ai-disruption-to-open-source-software-oss-377f10be2d8a)
- How can humans and AI agents best work together?
- Attribution and credit:
- How should we recognize contributions in an age of AI-assisted commits?
- Transparency: Should there be mandatory requirements to disclose AI use, including models and prompts used?
- Human ownership: should authors always remain legally and ethically accountable for the outputs of their code?
- Licensing: do open source licenses still mean anything when coding agents can translate or reimplement code?
- Sustainability: who does the long-term hard work of maintaining open source projects when AI does the “easy” work?
- Credit for training data: part of what AI proposes is reusing existing human-coded work without crediting it. Can there be a way to fairly credit the contribution of an open source project to the (often non open-source) models?
- When AI is the user: should open source projects be designed for machine consumers?
- The deadly feedback loop: models are trained on what they produce. Does this really work?
- Open data in the AI era: balancing access with protection from misuse
We look forward to seeing your thoughts on these topics! Please be sure to submit your abstract by April 9 if you want to be considered for a talk.