Academic journal publishing has been engaged in an open access reckoning for two decades. Conference proceedings — the primary publication format for computer science, engineering, and a growing number of other fields — are now entering the same debate. The question is the same: who should pay for access to publicly funded research? The answers are messier than the journal world's ongoing resolution suggests.
The Current State: A Patchwork of Models
Conference proceedings exist in a fragmented landscape of access models:
- IEEE Xplore: The majority of IEEE conference papers are paywalled — accessible to IEEE members and subscribers, or individually purchasable. Author self-archiving of accepted manuscripts is permitted, but the IEEE-formatted version requires a subscription.
- ACM Digital Library: Following ACM Open agreements with many universities, an increasing share of ACM conference papers are open access. Authors at non-member institutions may face an APC (article processing charge) for Gold OA.
- Springer LNCS / CCIS: Springer conference volumes are typically paywalled; open access is available at significant APC cost (often €1,500–€2,000 per paper).
- USENIX: Fully open access by default. All USENIX technical papers (SOSP, OSDI, USENIX Security, NSDI, etc.) are freely available on the USENIX website. No APC required. This model is financially sustained by registration fees and industry sponsorship.
- arXiv-first communities: In ML and AI, papers are routinely posted to arXiv before, during, or after conference acceptance. The majority of NeurIPS, ICML, and ICLR papers are freely accessible via arXiv even when the official proceedings are publisher-controlled.
- Open-access proceedings platforms: JMLR's Proceedings of Machine Learning Research (PMLR) hosts proceedings for many AI/ML workshops and conferences at no cost and with full open access. This model has become standard for the ICML, AISTATS, and UAI communities.
The Case for Open Access Proceedings
The strongest argument for open access is epistemic: publicly funded research should be publicly accessible. In most countries, university research is substantially government-funded. Placing that research behind paywalls accessible only to well-funded institutions creates a two-tier system where researchers at less wealthy institutions — particularly in the Global South — cannot access the literature that defines their field.
Practical benefits include:
- Higher citation rates — studies consistently show open access papers receive more citations across all fields
- Greater reach beyond academia — practitioners, policymakers, and journalists can access findings without an institutional subscription
- Preservation — papers indexed in open repositories are more likely to survive publisher changes, mergers, and discontinuations
The Case for Traditional Publishing Models
The counterargument is financial. High-quality peer review, proceedings production, archiving, and indexing cost real money. The current model transfers that cost to institutional subscribers rather than authors — which benefits researchers at well-funded institutions who pay nothing at the point of publication.
The APC model (Gold OA) shifts costs to authors — or more precisely, to research grants and universities. For researchers without grant funding or at institutions that do not cover APCs, this model can make publication prohibitively expensive.
The arXiv Effect: A De Facto Solution
In fields where arXiv preprints are standard (CS, AI, physics, mathematics, economics), the open access debate is somewhat moot in practice. The accepted manuscript version of nearly every significant CS conference paper is freely available on arXiv within days of acceptance — sometimes before the conference itself.
This creates an interesting situation: the official proceedings version (with DOI, official page numbers, and publisher branding) is paywalled, while the substantively identical preprint is openly accessible. Most researchers access the preprint; institutions pay for the official version for archiving purposes.
Funding Body Mandates: The Accelerant
Open access mandates from major funders are forcing a faster transition than the community would naturally choose:
- UK Research and Innovation (UKRI) requires all UKRI-funded papers to be open access, including conference papers, within 12 months of publication
- The European Commission's Horizon Europe programme has similar requirements
- In the US, the 2022 OSTP memorandum requires federal agencies to implement zero-embargo open access policies for funded research by 2025
This is driving rapid adoption of ACM Open institutional agreements and pushing more conferences toward PMLR or similar open models.
What This Means for Conference Organisers
Organisers choosing a publication model in 2026 should consider:
- Where is your likely author base? If predominantly EU or UK-funded, open access compliance is a requirement, not a choice
- Is the PMLR or similar model viable? For CS and AI conferences, open access proceedings have become the community expectation in many subfields
- What APC level can your authors realistically bear? Charging €2,000 per paper for open access effectively excludes unfunded researchers
What This Means for Authors
Check your funder's open access policy before selecting a venue. Many researchers discover OA requirements only after acceptance — at which point their options are limited to paying an APC or posting a preprint and hoping the embargo period satisfies their funder's requirements.
The safest strategy: always post your accepted manuscript to arXiv or your institutional repository, regardless of the conference's model. This satisfies most funder mandates, increases your paper's reach, and ensures it remains accessible regardless of what happens to the publisher.