Reimagining Open Source and Openness in AI: Co-Creating Responsible Technological Futures

May combine two claims.

The current AI field is shaped by rapid cycles of innovation and hype, with dominant generative AI narratives and agendas largely led by a narrow set of powerful actors who exercise disproportionate influence in how openness is defined, operationalized, and legitimized.

“The current AI field is shaped by rapid cycles of innovation and hype, with dominant generative AI narratives and agendas largely led by a narrow set of powerful actors (particularly large technology firms, but also influential standards bodies and platform owners) who exercise disproportionate influence in how openness is defined, operationalized, and legitimized.”

• cites [6] [6] Benjamin, R. Race After Technology: Abolitionist Tools for the New Jim Code . Polity, Cambridge, UK, 2019. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [8] [8] Birhane, A. Algorithmic injustice: a relational ethics approach. Patterns 2 , 2 (2021). - sourcePartly supported

  Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured
• cites [21] [21] Crawford, K. The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence . Yale University Press, New Haven, CT, USA, 2022. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

Claims of openness can amount to 'open washing', where selective disclosure results in reputational benefits without meaningful reproducibility, accountability or redistribution of power.

“scholars have warned that claims of openness can amount to 'open washing', where selective disclosure results in reputational benefits without meaningful reproducibility, accountability or redistribution of power [36].”

• cites [36] [36] Liesenfeld, A., and Dingemanse, M. Rethinking open source generative AI: open-washing and the EU AI act. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency (2024), ACM. - sourceSupported

May combine two claims.

This concentration of agenda-setting power and dominance of a small set of corporate actors risks narrowing the imaginaries of AI futures, privileging commercial interests over broader social concerns.

“This concentration of agenda-setting power and dominance of a small set of corporate actors risks narrowing the imaginaries of AI futures, privileging commercial interests over broader social concerns [6, 8, 21].”

• cites [6] [6] Benjamin, R. Race After Technology: Abolitionist Tools for the New Jim Code . Polity, Cambridge, UK, 2019. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [8] [8] Birhane, A. Algorithmic injustice: a relational ethics approach. Patterns 2 , 2 (2021). - sourcePartly supported

  In “Knowing that centers human relations” (page unavailable), Birhane argues that computing and data science are dominated by privileged groups, limiting the perspectives and concerns reflected in their work. The paper does not establish that a small set of corporate actors controls AI agenda-setting or that commercial interests are privileged over broader social concerns. (LLM, conf=0.84)

  Experimental — accuracy not yet measured
• cites [21] [21] Crawford, K. The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence . Yale University Press, New Haven, CT, USA, 2022. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Openness alone does not guarantee accountability, equity, or democratic control, while models labeled as 'open' can reproduce structural inequities or obscure critical aspects of their creation and governance.

“Openness alone does not guarantee accountability, equity, or democratic control, while models labeled as 'open' can reproduce structural inequities or obscure critical aspects of their creation and governance [63].”

• cites [63] [63] Widder, D. G., Whittaker, M., and West, S. M. Why 'open' AI systems are actually closed, and why this matters. Nature 635 , 8040 (2024), 827-833. - sourceSupported

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

Open source and open AI models promise to advance research, innovation, transparency, and equitable participation.

“Open source and open AI models promise to advance research, innovation, transparency, and equitable participation [24] [50].”

• cites [24] [24] Eiras, F., Petrov, A., Vidgen, B., Schroeder, C., Pizzati, F., Elkins, K., Mukhopadhyay, S., Bibi, A., Purewal, A., Botos, C., Steibel, F., Keshtkar, F., Barez, F., Smith, G., Guadagni, G., Chun, J., Cabot, J., Imperial, J., Nolazco, J. A., Landay, L., Jackson, M., Torr, P. H. S., Darrell, T., Lee, Y., and Foerster, J. Risks and opportunities of open-source generative ai, 2024. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [50] [50] Seger, E., Dreksler, N., Moulange, R., Dardaman, E., Schuett, J., Wei, K., Winter, C., Arnold, M., Ó hÉigeartaigh, S., Korinek, A., Anderljung, M., Bucknall, B., Chan, A., Stafford, E., Koessler, L., Ovadya, A., Garfinkel, B., Bluemke, E., Aird, M., Levermore, P., Hazell, J., and Gupta, A. Open-sourcing highly capable foundation models: An evaluation of risks, benefits, and alternative methods for pursuing open-source objectives. SSRN Electronic Journal (2023). - sourceSupported

Classified as probably not a substantial claim.

'Open' narratives often obscure who can meaningfully participate.

“This echoes [47] arguing openness should be understood at the level of AI ecosystems as opposed to individual models, and [63] noting that 'open' narratives often obscure who can meaningfully participate.”

• cites [63] [63] Widder, D. G., Whittaker, M., and West, S. M. Why 'open' AI systems are actually closed, and why this matters. Nature 635 , 8040 (2024), 827-833. - sourceSupported

Probably not a substantial claim: may not pass S2 (checkable).

Openness should be understood at the level of AI ecosystems as opposed to individual models.

“This echoes [47] arguing openness should be understood at the level of AI ecosystems as opposed to individual models, and [63] noting that 'open' narratives often obscure who can meaningfully participate.”

• cites [47] [47] Paris, T., Moon, A., and Guo, J. L. Opening the scope of openness in ai. In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency (FAccT '25) (2025), Association for Computing Machinery. - sourceSupported
• cites [63] [63] Widder, D. G., Whittaker, M., and West, S. M. Why 'open' AI systems are actually closed, and why this matters. Nature 635 , 8040 (2024), 827-833. - sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing).

Prior research has illustrated that companies embrace 'open' AI and use rhetoric of 'open' to expand market power.

“While prior research has illustrated that companies embrace 'open' AI and use rhetoric of 'open' to expand market power [63], our findings reveal that the openness spectrum itself affords companies a unique form of structural defensibility because legitimate value frameworks exist to justify varying degrees of openness.”

• cites [63] [63] Widder, D. G., Whittaker, M., and West, S. M. Why 'open' AI systems are actually closed, and why this matters. Nature 635 , 8040 (2024), 827-833. - sourceSupported

Classified as probably not a substantial claim.

The workshop generated action pathways and a research roadmap toward responsible openness in AI across eight cross-cutting dimensions.

“the workshop generated concrete action pathways and a research roadmap toward responsible openness in AI across eight cross-cutting dimensions”

Classified as probably not a substantial claim. May combine two claims.

Research priorities included examining AI training and education models, labor and recognition structures in open source development, and interdisciplinary collaborations that support sustained participation.

“Research priorities included examining AI training and education models, labor and recognition structures in open source development, and interdisciplinary collaborations that support sustained participation.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Participants envisioned a world of shared prosperity where all people have resources and opportunities to live dignified, fulfilling lives.

“Participants envisioned a world of shared prosperity, whereby all people have the resources and opportunities to live dignified, fulfilling lives.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Community and human connection emphasized designing AI systems that strengthen social bonds and collective resilience rather than replacing or distorting human relationships.

“Community and human connection emphasized designing AI systems that strengthen social bonds and collective resilience rather than replacing or distorting human relationships.”

Classified as probably not a substantial claim.

A recurring tension concerned balancing the expansion of openness with the need to ensure meaningful access and safe use across diverse contexts.

“A recurring tension concerned balancing the expansion of openness with the need to ensure meaningful access and safe use across diverse contexts.”

May combine two claims.

Participants broadly converged on shared aspirations during visioning, but greater divergence emerged on how openness should function in practice and what should be prioritized.

“While participants broadly converged on shared aspirations during the visioning exercise, greater divergence emerged as they considered how openness should function in practice and what should be prioritized.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The findings trace a progression from shared aspirations to concrete proposals and unresolved tensions.

“The findings trace a progression from shared aspirations to concrete proposals and unresolved tensions.”

Classified as probably not a substantial claim.

Transparency was framed as reciprocal accountability among developers, institutions, and the public, rather than one-way disclosure.

“Transparency was framed less as one-way disclosure, and more as reciprocal accountability among developers, institutions, and the public.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The workshop design surfaced both common ground and tensions.

“the design surfaced both common ground and tensions.”

Classified as probably not a substantial claim. May combine two claims.

Central to the envisioned future was a commitment to social and economic justice, including respect for human rights, redistribution of power, and conditions for collective wellbeing.

“Central to this vision was a commitment to social and economic justice, including respect for human rights, redistribution of power, and conditions for collective wellbeing.”

Classified as probably not a substantial claim.

Openness was framed as shared stewardship in which AI systems are treated as digital commons for collective benefit, not simply as a technical attribute of model release.

“Openness was framed not simply as a technical attribute of model release, but as a form of shared stewardship in which AI systems are treated as digital commons for collective benefit.”

Probably not a substantial claim: may not pass S1 (asserted), S2 (checkable), S3 (load-bearing). May combine two claims.

Additional research priorities include mapping interlinkages between OSS, open data, open models, and open science communities, and studying how operational practices shape meaningful openness beyond formal licensing.

“Additional priorities include mapping interlinkages between OSS, open data, open models, and open science communities; and studying how operational practices (e.g., governance, contribution management) shape meaningful openness beyond formal licensing.”

Classified as probably not a substantial claim.

Participants identified openness and transparency as foundational conditions for realizing and sustaining such a world.

“Participants identified openness and transparency as foundational conditions for realizing and sustaining such a world.”

Classified as probably not a substantial claim.

A tension emerged regarding whether openness should function as a compulsory baseline or be conditional.

“exposing a tension regarding whether openness should function as a compulsory baseline or be conditional.”

Probably not a substantial claim: may not pass S1 (asserted). May combine two claims.

Research priorities included clarifying how openness can be defined and measured, developing frameworks for gradients of openness, and examining stakeholder understandings of openness.

“Research priorities focused on clarifying how openness can be defined and measured in practice, including developing frameworks to characterize and measure gradients of openness across the AI stack, as well as examining how different stakeholders (e.g., communities, companies and governments) understand and value openness.”

Classified as probably not a substantial claim. May combine two claims.

Proposed actions included building public or shared compute infrastructure, supporting 'public option' AI systems, creating experimentation sandboxes, and developing shared hardware and tooling to reduce dependence on dominant infrastructure providers.

“Proposed actions included building public or shared compute infrastructure, supporting 'public option' AI systems, creating experimentation sandboxes for open models and datasets, and developing shared hardware and tooling to reduce dependence on dominant infrastructure providers.”

Classified as probably not a substantial claim.

Evaluation and standards were framed as essential to making open AI systems reproducible, trustworthy, and legible.

“Evaluation and standards were framed as essential to making open AI systems reproducible, trustworthy, and legible.”

Probably not a substantial claim: may not pass S1 (asserted), S2 (checkable).

Research priorities focused on studying effective governance models that balance openness with accountability across public, private, and community actors.

“Research priorities focused on studying effective governance models that balance openness with accountability across public, private, and community actors.”

Classified as probably not a substantial claim. May combine two claims.

Action pathways included establishing participatory multistakeholder governance structures, promoting data privacy through federated approaches, enabling model unlearning or withdrawal, and strengthening oversight through procurement, liability, and accountability mechanisms.

“Action pathways included establishing participatory, multistakeholder governance structures for open AI models; promoting data privacy through federated approaches to data access and non-predatory data ecosystems that strengthen individual and community data rights; enabling model unlearning or withdrawal; and strengthening oversight through procurement, liability, and accountability mechanisms.”

Classified as probably not a substantial claim.

Participants emphasized moving beyond tokenistic consultation toward governance arrangements that grant communities real decisionmaking power.

“Participants emphasized moving beyond tokenistic consultation toward governance arrangements that grant communities real decisionmaking power.”

Classified as probably not a substantial claim. May combine two claims.

Research priorities included identifying safety, security, and risk-mitigation approaches specific to open models, clarifying which properties can be guaranteed in open settings, and developing safeguards against harmful and extractive use.

“Research priorities included identifying safety, security, and risk-mitigation approaches specific to open models; clarifying which properties (e.g., explainability, privacy) can realistically be guaranteed in open settings; and developing safeguards against harmful and extractive use.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Social and economic justice was often articulated as expanding equality of opportunity, with AI as a means to enable equitable outcomes, democratize access to resources, support human rights, and redistribute power.

“Social and economic justice was often articulated in terms of expanding equality of opportunity, with AI as a means to enable equitable outcomes, democratize access to resources, support human rights, and redistribute power.”

Classified as probably not a substantial claim. May combine two claims.

Action pathways included broadening participation beyond a small number of geopolitical and institutional actors by investing in AI literacy and education, shared technical and institutional infrastructure, and community-led model development.

“Action pathways included broadening participation beyond a small number of geopolitical and institutional actors by investing in AI literacy and education, shared technical and institutional infrastructure, and community-led model development.”

Classified as probably not a substantial claim. May combine two claims.

Public procurement was identified as a key lever, with proposals to require open data standards, model documentation, and environmental disclosures, making openness the norm for publicly-funded systems while allowing narrow exceptions when openness could endanger vulnerable groups.

“Public procurement was identified as a key lever, with proposals to require open data standards, model documentation, and environmental disclosures, making openness the norm for publicly-funded systems while allowing for narrow exceptions when openness could endanger vulnerable groups.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Action pathways focused on strengthening safety and security practices for open models, including safeguards against misuse, extractive deployment, and harmful downstream applications, while prioritizing fairness, transparency, and interpretability.

“Action pathways focused on strengthening safety and security practices for open models, including safeguards against misuse, extractive deployment, and harmful downstream applications; while also prioritizing fairness, transparency, and interpretability.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Key research areas included collective and participatory governance approaches, comparative analyses of regulatory and legal frameworks, and empirical studies of decision-making around openness.

“Key areas included collective and participatory governance approaches in open AI systems, comparative analyses of regulatory and legal frameworks, and empirical studies of decision-making around openness within governments, companies and large-scale open AI projects.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Research priorities centered on developing rigorous, socially relevant evaluation frameworks, including domain- and task-specific benchmarks, open evaluation ecosystems, and testing regimes beyond technical performance.

“Research priorities centered on developing rigorous, socially relevant evaluation frameworks, including domain- and task-specific benchmarks across the AI stack, open evaluation ecosystems, and testing regimes that extend beyond technical performance to capture safety, accountability, and societal impact.”

Probably not a substantial claim: may not pass S3 (load-bearing).

AI was imagined as a tool deliberately developed and governed to advance these goals.

“AI was imagined as a tool deliberately developed and governed to advance these goals.”

May combine two claims.

Many proposed actions were complementary or could coexist, while others surfaced tensions.

“While many proposed actions were complementary or could coexist, others surfaced tensions.”

Classified as probably not a substantial claim. May combine two claims.

Tensions emerged around accountability, liability, and how community governance, including context-dependent openness, can coexist with open source norms.

“Tensions emerged around accountability, liability, and how community governance, including context-dependent openness, can coexist (or not) with open source norms.”

Classified as probably not a substantial claim. May combine two claims.

Key research topics for usability include making open models more widely deployable and accessible, while improving transparency and explainability for public understanding.

“Key research topics for usability include how to make open models more widely deployable and accessible to different people (e.g., locally runnable, multilingual, disability-inclusive, adaptable), while improving transparency and explainability for public understanding.”

Classified as probably not a substantial claim.

Participants emphasized that openness alone does not guarantee responsibility.

“Participants emphasized that openness alone does not guarantee responsibility.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Several breakout groups called to prioritize smaller, more sustainable models over endless scaling.

“Several breakout groups included calls to prioritize smaller, more sustainable models over endless scaling.”

Classified as probably not a substantial claim.

Participants emphasized that meaningful collaboration requires openness across the AI stack combined with interoperability and mechanisms for collective input.

“Participants emphasized that meaningful collaboration requires openness across the AI stack (i.e. data, models, software, and infrastructure) combined with interoperability and mechanisms for collective input.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Many proposed actions and research directions were complementary, while others surfaced tensions where approaches could not easily coexist.

“Many proposed actions and research directions were complementary, while others surfaced tensions where approaches could not easily coexist.”

Classified as probably not a substantial claim.

Participants emphasized that access to compute, tooling, and shared resources shapes who can build, evaluate, and contest open AI systems.

“Participants emphasized that access to compute, tooling, and shared resources shapes who can build, evaluate, and contest open AI systems.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Democratic governance was framed around decentralizing power in AI development and oversight, with participatory stewardship and trusted public institutions acting in the public interest.

“Democratic governance was framed around decentralizing power in AI development and oversight, with participatory stewardship and trusted public institutions acting in the public interest.”

Classified as probably not a substantial claim. May combine two claims.

Governance discussions centered on who has authority and responsibility over open AI systems, how decisions are made and enforced, and how power is distributed.

“Governance discussions centered on who has authority and responsibility over open AI systems, how decisions are made and enforced, and how power is distributed.”

Classified as probably not a substantial claim.

Participants framed openness as operating along gradients, with different implications across data, models, software, and infrastructure.

“Participants framed openness as operating along gradients, with different implications across data, models, software, and infrastructure.”

Classified as probably not a substantial claim.

Participants stressed that without attention to capacity and labor, open AI risks reproducing existing inequalities even when artifacts are publicly available.

“Participants stressed that without attention to capacity and labor, open AI risks reproducing existing inequalities even when artifacts are publicly available.”

Classified as probably not a substantial claim.

Participants emphasized that openness without robust evaluation risks becoming symbolic rather than accountable.

“Participants emphasized that openness without robust evaluation risks becoming symbolic rather than accountable.”

Classified as probably not a substantial claim.

Inclusive design and capacity building were framed as conditions for responsible openness rather than a downstream benefit, focusing on who can participate and under what conditions.

“Inclusive design and capacity building were framed as conditions for responsible openness rather than a downstream benefit, focusing on who is able to participate in design and development of open AI systems and under what conditions.”

Classified as probably not a substantial claim. May combine two claims.

Responsibility was framed as preventing harm and ensuring open AI systems are safe, usable, and meaningful in practice.

“Responsibility was framed as preventing harm and ensuring open AI systems are safe, usable, and meaningful in practice.”

Classified as probably not a substantial claim.

Participants highlighted that openness is insufficient if communities lack the capacity to meaningfully participate in design, development, and use.

“Participants highlighted that openness is insufficient if communities lack the capacity to meaningfully participate in design, development, and use.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Sustainability was invoked in relation to green AI supply chains and aligning technological progress with planetary well-being.

“Finally, sustainability was invoked in relation to green AI supply chains and aligning technological progress with planetary well-being.”

May combine two claims.

The workshop generated co-created visions of desirable futures and the role of AI, alongside a set of action pathways and a research roadmap focused on responsible open source and openness in AI.

“The workshop generated co-created visions of desirable futures and the role of AI, alongside a set of action pathways and a research roadmap focused on responsible open source and openness in AI.”

• AssumptionRests on an assumption: There has been limited empirical work examining how diverse stakeholders collectively understand and negotiate responsible openness in AI, particularly through participatory processes that extend beyond industry-led definitions and frameworks.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

This balancing introduced tension, but also enabled cross-disciplinary thinking and generated higher-level reflections on values and governance.

“This introduced tension, but also enabled cross-disciplinary thinking and generated higher-level reflections on values and governance.”

May combine two claims.

In many cases these different perspectives and priorities could converge, whereas in other cases they resulted in contradictions.

“In many cases these different perspectives and priorities could converge, whereas in other cases they resulted in contradictions.”

Probably not a substantial claim: may not pass S3 (load-bearing).

One orientation imagined openness as primarily a production task.

“In this, one orientation imagined openness as primarily a production task (e.g., creating and releasing more open models and datasets).”

Classified as probably not a substantial claim.

Participants emphasized that openness should be considered and evaluated through its application and impact, not solely through artifact availability.

“Participants emphasized that openness should be considered and evaluated through its application and impact, not solely through artifact availability.”

May combine two claims.

Participants highlighted limits of treating openness as primarily artifact release, a framing more common in traditional OSS, and emphasized participatory structures that redistribute decision-making alongside technical openness.

“Participants highlighted limits of treating openness as primarily artifact release, a framing that is more common in traditional OSS, and emphasized participatory structures that redistribute decision-making alongside technical openness.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Action pathways included developing stable public and pooled funding mechanisms, compensating contributors, and creating incentives that reward safe, responsible, and environmentally conscious development.

“Action pathways included developing stable public and pooled funding mechanisms, compensating contributors, and creating incentives that reward safe, responsible, and environmentally conscious development.”

Probably not a substantial claim: may not pass S3 (load-bearing).

At times, groups had to balance depth of discussion with inclusivity, occasionally limiting technical specificity to maintain collective engagement.

“At times, groups had to balance depth of discussion with inclusivity, occasionally limiting technical specificity to maintain collective engagement.”

Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing). May combine two claims.

Research priorities included comparative studies of open versus closed approaches in different domains and contexts, and assessments and historical analyses of past technologies to understand how openness, governance, and public investment have influenced shared prosperity.

“Research priorities included comparative studies of open versus closed approaches in different domains and contexts, assessments and historical analyses of past technologies to understand how openness, governance, and public investment have influenced shared prosperity.”

Classified as probably not a substantial claim.

Action pathways focused on supporting open AI models and datasets for application in civic technology, climate and environmental modeling, healthcare, scientific research, and public service delivery, often favoring smaller or purpose-built models.

“Action pathways focused on supporting open AI models and datasets for application in civic technology, climate and environmental modeling, healthcare, scientific research, and public service delivery, often calling for and favoring smaller or purpose-built models.”

May combine two claims.

Tensions appeared most clearly in the action planning session but surfaced throughout.

“They appeared most clearly in the action planning session, but surfaced throughout including in group discussions and share-outs.”

Classified as probably not a substantial claim.

Several workshop participants emphasized that openness does not automatically produce inclusive development or governance.

“Several participants emphasized that openness does not automatically produce inclusive development or governance.”

Classified as probably not a substantial claim.

This tension centers on whether openness alone translates to real utility and access.

“This centers less on the purpose of openness and more on whether openness alone translates to real utility and access.”

Classified as probably not a substantial claim.

Unresolved questions about openness surfaced in workshop discussions about building open foundation model ecosystems that are powerful enough to rival closed AI labs and genuinely decentralized, while not replicating the extractive dynamics they seek to counter.

“These unresolved questions surfaced in workshop discussions about building open foundation model ecosystems that are powerful enough to rival closed AI labs and genuinely decentralized, while not replicating the extractive dynamics they seek to counter.”

Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

Creating pre-read materials was critical for ensuring shared understanding across participants with varying technical expertise.

“Creating pre-read materials was critical.”

Probably not a substantial claim: may not pass S3 (load-bearing).

One action emphasized not imposing higher expectations on open than on closed models.

“One action emphasized, 'Not imposing higher expectations (ethical, responsible) on open than [on] closed models.'”

Probably not a substantial claim: may not pass S3 (load-bearing).

This tension highlights differing expectations about whether responsible openness is primarily about making things open or making openness work for different people.

“This tension highlights differing expectations about whether responsible openness is primarily about making things open or making openness work for different people.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Environmental and social impacts were highlighted as cross-cutting concerns and areas for action and research, particularly as openness can increase reuse and deployment.

“Environmental and social impacts were highlighted as cross-cutting concerns and areas for action and research, particularly as openness can increase reuse and deployment.”

• AssumptionRests on an assumption: Openness can increase reuse and deployment.

Probably not a substantial claim: may not pass S3 (load-bearing).

Some contributions treated openness itself as the desired end state.

“Some contributions treated openness itself as the desired end state-as a value or principle to maximize in its own right.”

May combine two claims.

Some inputs emphasized liability frameworks and accountability mechanisms for harms associated with open systems, while others called for legal protections or safe harbors for developers releasing open models.

“Some inputs emphasized liability frameworks and accountability mechanisms for harms associated with open systems (aligned with openness as a means), while others called for legal protections or safe harbors to ensure that developers were not potentially penalized or faced repercussions for releasing open models (aligned with openness as an end).”

Probably not a substantial claim: may not pass S3 (load-bearing).

Some participants and breakout groups shared actions whereby openness and open-source is considered mandatory.

“Some participants and breakout groups shared actions whereby openness and open-source is considered mandatory.”

Probably not a substantial claim: may not pass S3 (load-bearing).

This tension intersected with discussions of legal and accountability structures.

“This tension intersected with discussions of legal and accountability structures.”

Probably not a substantial claim: may not pass S3 (load-bearing).

This tension is reflected in two seemingly contradictory actions: 'Safe harbor for open developers' versus 'Liability regime that holds developers accountable for foreseeable harms'.

“This tension is reflected in two seemingly contradictory actions, 'Safe harbor for open developers' versus 'Liability regime that holds developers accountable for foreseeable harms'.”

Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing).

One participant noted that attention on societal questions that aren't just technical may be better as an alternative approach to corporate, closed models.

“As one participant noted, 'attention on societal questions that aren't just technical... may be better as an alternative approach [to corporate, closed models].'”

May combine two claims.

The workshop revealed that a single, unified responsible open source AI future was neither clearly articulated across breakout groups nor necessarily desirable.

“The workshop revealed that a single, unified responsible open source AI future was neither clearly articulated across breakout groups nor necessarily desirable.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A recurring challenge in the workshop was ensuring shared understanding across participants with varying technical expertise.

“A recurring challenge was ensuring shared understanding across participants with varying technical expertise.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Illustrative cases of public investment in open source infrastructure were identified as valuable empirical sites for understanding how policy can support digital commons beyond large foundation models.

“Illustrative cases of public investment in open source infrastructure were identified as valuable empirical sites for understanding how policy can support digital commons beyond large foundation models.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The participatory process surfaced several methodological tensions that are informative for futures and participatory design research.

“The participatory process surfaced several methodological tensions that are informative for futures and participatory design research.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In some cases, participants discussed how open source models and data can embody additional values and attributes, such as being ethically superior to closed models.

“In some cases, participants discussed how open source models and data can embody additional values and attributes, such as being ethically superior to closed models.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

One group's actions captured the contradiction explicitly: 'Openness is not voluntary; it is compulsory,' alongside 'Closed when necessary to protect socially marginalized communities.'

“Actions developed in one group capture the contradiction explicitly: 'Openness is not voluntary; it is compulsory,' alongside 'Closed when necessary to protect socially marginalized communities.'”

Classified as probably not a substantial claim. May combine two claims.

Another orientation emphasized that responsible openness requires meaningful access, usability across diverse contexts, and forms of participation and engagement.

“Another emphasized that responsible openness requires meaningful access, usability across diverse contexts, and forms of participation and engagement.”

Classified as probably not a substantial claim.

Others emphasized model openness as a spectrum, with responsible openness requiring selective limitation or conditions.

“Meanwhile, others emphasized model openness as a spectrum, with responsible openness requiring selective limitation or conditions.”

May combine two claims.

This tension extended to datasets, with some actions calling for open data to support reproducibility and inclusion, and others highlighting how openness can reproduce extractive dynamics.

“This tension extended to datasets, with some actions calling for open data to support reproducibility and inclusion (e.g., 'open data from the global majority [that could facilitate models that work for more people]'), and others highlighting how openness can reproduce extractive dynamics.”

Classified as probably not a substantial claim. May combine two claims.

Others held openness as meaningful when it translated into real capability and participation, prioritizing usability, meaningful access and use across communities, reflecting values of utility, inclusion, and practical benefit.

“Others held openness as meaningful when it translated into real capability and participation, prioritizing usability, meaningful access and use across communities-reflecting values of utility, inclusion, and practical benefit.”

May combine two claims.

For some participants, openness itself was the primary ethical and political commitment, with actions emphasizing openness as an end in its own right, focusing on expanding openness and framing open source as mandatory, reflecting values tied to freedom, transparency, and distribution of power.

“For some, openness itself was the primary ethical and political commitment: actions emphasized openness as an end in its own right, focusing on expanding openness and framing open source as mandatory-reflecting values tied to freedom, transparency, and distribution of power.”

Classified as probably not a substantial claim.

One workshop action called for diverse participants in development/deployment globally to develop open source AI models responsive to their needs and interests, emphasizing participation, stewardship, and community governance as central for responsible openness.

“One action called for 'diverse participants in development/deployment [globally] to develop open source AI models that are responsive to their needs and interests,' emphasizing participation, stewardship, and community governance as central for responsible openness.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Others cautioned against or included explicit actions to not have higher standards for open source models and datasets, as heightened expectations could create new burdens, disincentivize contribution, or slow development.

“Others cautioned against or included explicit actions to not have higher standards for open source models and datasets, as heightened expectations could create new burdens, disincentivize contribution, or slow development.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Some workshop contributions articulated a broader commons-oriented view, treating openness as the primary mechanism through which collective governance and inclusion could emerge.

“At the same time, some contributions articulated a broader commons-oriented view, treating openness as the primary mechanism through which collective governance and inclusion could emerge.”

Probably not a substantial claim: may not pass S1 (asserted). May combine two claims.

A final tension in the workshop was whether openness is sufficient on its own or whether its value depends fundamentally on how open data and models are developed, governed, and used, and by whom.

“A final tension concerned whether openness is sufficient on its own or whether its value depends fundamentally on how open data and models are developed, governed, and used-and by whom.”

Classified as probably not a substantial claim. May combine two claims.

And others surfaced openness as a means rather than an inherent good, emphasizing conditional openness, protections, and governance structures centering marginalized communities, reflecting values of equity and justice.

“And others surfaced openness as a means rather than an inherent good, emphasizing conditional openness, protections, and governance structures centering marginalized communities-reflecting values of equity and justice.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Others framed openness as instrumentally valuable when it supports broader social aims.

“Others framed openness as instrumentally valuable when it supports broader social aims (e.g., equity, community benefit, shifting power hierarchies).”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

While these two orientations are not inherently incompatible, they reflect different priorities and resource commitments.

“While these two orientations are not inherently incompatible, they reflect different priorities and resource commitments.”

Classified as probably not a substantial claim.

Participants stressed that responsible openness requires sustainable economic conditions.

“Participants stressed that responsible openness requires sustainable economic conditions.”

May combine two claims.

Some participants emphasized full transparency to advance accountability and innovation, while other participants called for conditional openness of models to support community needs for sovereignty or safety.

“Some participants emphasized full transparency to advance accountability and innovation, while other participants called for conditional openness of models to support community needs for sovereignty or safety.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

One group's actions highlighted positive aspects of openness such as being more auditable and transparent, while also adding 'open = inclusive... open = unbiased.'

“This is reflected in one group's actions that highlighted positive aspects of openness such as being more auditable and transparent, while also adding 'open = inclusive... open = unbiased.'”

Probably not a substantial claim: may not pass S1 (asserted), S2 (checkable).

Research priorities focused on examining infrastructural conditions under which openness in AI can be equitable and sustainable, including public compute infrastructure, cross-government data-sharing, and technical approaches to reducing infrastructure demands.

“Research priorities focused on examining infrastructural conditions under which openness in AI can be equitable and sustainable, including public compute infrastructure, cross-government data-sharing, and technical approaches to reducing infrastructure demands (e.g., distillation, quantization).”

May combine two claims.

Our workshop surfaced co-created visions, action pathways, and research priorities alongside productive tensions revealing that openness is an ongoing sociotechnical project informed by values, priorities, and positionalities.

“Our workshop surfaced co-created visions, action pathways, and research priorities-alongside productive tensions revealing that openness is an ongoing sociotechnical project informed by values, priorities, and positionalities.”

Classified as probably not a substantial claim.

Without long-term funding and fair attribution, open AI ecosystems risk fragility and capture.

“Without long-term funding and fair attribution, open AI ecosystems risk fragility and capture.”

May combine two claims.

Linked to these concerns were strong themes of consent, privacy, and protection, including recognition that some data, knowledge, or systems should not be fully open due to cultural, ethical, or safety considerations.

“Linked to these concerns were strong themes of consent, privacy, and protection, including recognition that some data, knowledge, or systems should not be fully open due to cultural, ethical, or safety considerations.”

May combine two claims.

While there was broad alignment that openness is key to redistributing power away from dominant actors, contradictions emerged around how much openness is desirable and when it should be constrained.

“While there was broad alignment that openness is key to redistributing power away from dominant actors, contradictions emerged around how much openness is desirable and when it should be constrained (if at all).”

Probably not a substantial claim: may not pass S3 (load-bearing).

Another tension concerned whether openness in itself is sufficient.

“Another tension concerned whether openness in itself is sufficient.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Some emphasized a 'small is beautiful' approach, valuing purpose-built, contextually appropriate open systems over large-scale openness.

“Relatedly, some emphasized a 'small is beautiful' approach, valuing purpose-built, contextually appropriate open systems that are driven by particular purpose and use cases, over large-scale openness.”

Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing). May combine two claims.

This perspective is tied to views that openness can inherently advance broader societal aims more effectively than closed models, and that ethical development can serve as a differentiator from closed models.

“This perspective is tied to views that openness can inherently advance broader societal aims (e.g., inclusion) more effectively than closed models, and that ethical development can serve as a differentiator from closed models.”

Classified as probably not a substantial claim.

Participants emphasized that genuine access requires education and support structures that enable different communities to meaningfully engage.

“Participants emphasized that genuine access requires education and support structures that enable different communities to meaningfully engage.”

Probably not a substantial claim: may not pass S1 (asserted), S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

Research priorities focused on comparing funding models for open AI, assessing economic viability over time, studying incentive structures for participation and maintenance, and examining how costs and benefits are distributed across stakeholders.

“Research priorities focused on comparing funding models for open AI, assessing economic viability over time, studying incentive structures for participation and maintenance, and examining how costs and benefits are distributed across stakeholders.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Participant affiliations were academia 54.67%, industry 14.67%, civil society/nonprofit 20%, government 9.33%, foundations 1.33%.

“Participants' primary affiliations included academia (54.67%), industry (14.67%), civil society or nonprofit institutions (20%), government (9.33%), and foundations (1.33%).”

Classified as probably not a substantial claim.

Technology is not neutral, but rather shaped by the values, experiences, knowledge, and positionalities of its developers.

“this scholarship emphasizes that technology is not neutral, but rather shaped by the values, experiences, knowledge, and positionalities of its developers.”

• cites [20] [20] Costanza-Chock, S. Design Justice: Community-Led Practices to Build the Worlds We Need . MIT Press, Cambridge, MA, USA, 2020. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [23] [23] Dunne, A., and Raby, F. Speculative Everything: Design, Fiction, and Social Dreaming . MIT Press, Cambridge, MA, USA, 2013. - sourceNot checked — no full text of the source
• cites [27] [27] Haraway, D. Situated knowledges: The science question in feminism and the privilege of partial perspective. Feminist Studies 14 , 3 (1988), 575-599. - sourceNot checked — no full text of the source
• cites [28] [28] Harding, S. G. Whose Science? Whose Knowledge? Thinking from Women's Lives . Cornell University Press, Ithaca, NY, USA, 1991. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

Software licenses have long been understood not merely as legal instruments, but as expressions of the values their authors seek to uphold.

“Software licenses have long been understood not merely as legal instruments, but as expressions of the values their authors seek to uphold [44].”

• cites [44] [44] Open Source Initiative. Open-Source AI Definition 1.0. Open Source Initiative, October 2024. Accessed: 2025-12-15. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

Futures-oriented approaches extend participatory design by inviting communities to imagine and deliberate on alternative technological futures.

“Futures-oriented approaches extend participatory design by inviting communities to imagine and deliberate on alternative technological futures.”

• AssumptionRests on an assumption: Participatory and co-design methods have been central to the effort of envisioning alternative technological futures.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

These methods foreground plurality and productive tension rather than consensus, highlighting the role of 'contamination' in collective world-building.

“These methods foreground plurality and productive tension rather than consensus, highlighting the role of 'contamination', the process through which distinct entities become entangled, in collective world-building [39].”

• cites [39] [39] Lu, A. J., Wikstrom, E. V., and Dillahunt, T. R. Contamination, otherness, and negotiating bottom-up sociotechnical imaginaries in participatory speculative design. In Proceedings of the 2024 ACM Designing Interactive Systems Conference (New York, NY, USA, 2024), DIS '24, Association for Computing Machinery, pp. 1308-1324. - sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The primary coder also had a lead role in workshop facilitation and steering committee.

“Primary coding and synthesis were led by a researcher who also played a lead role in the workshop's facilitation and steering committee.”

Probably not a substantial claim: may not pass S4 (based).

Platforms such as PyTorch illustrate how infrastructural control translates into agenda-setting power by enabling owners to shape default architectures, workflows, and benchmarks.

“Platforms such as PyTorch further illustrate how infrastructural control translates into agenda-setting power by enabling owners to shape default architectures, workflows, and benchmarks.”

Classified as probably not a substantial claim.

Methods of Futures Workshops and Participatory Speculative Design draw on structured facilitation of collective imagination and critique to support visioning and pathway building.

“Methods of Futures Workshops [60] and Participatory Speculative Design draw on structured facilitation of collective imagination and critique to support visioning and pathway building.”

• cites [60] [60] Vidal, R. V. V. The future workshop: Democratic problem solving. Technical Report IMM-REP-2005-11, Informatics and Mathematical Modelling, Technical University of Denmark, Lyngby, Denmark, 2005. -Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The Open Source Initiative's Open Source AI Definition (OSAID) has emerged as an influential reference point for open source AI.

“the Open Source Initiative's Open Source AI Definition (OSAID) has emerged as an influential reference point”

• cites [43] [43] Open Source Initiative. The open source AI definition, 2024. Accessed 2026-01-11. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Open or open source AI models are frequently equated with open weights, yet weights alone do not ensure full transparency, reproducibility, or accountability when other critical components remain closed.

“Open or open source AI models are frequently equated with open weights, yet weights alone do not ensure full transparency, reproducibility, or accountability when other critical components remain closed.”

Probably not a substantial claim: may not pass S4 (based).

These systems are not neutral, as even openly available models embed particular priorities and perspectives shaped by their development contexts.

“At the same time, these systems are not neutral, as even openly available models embed particular priorities and perspectives shaped by their development contexts.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

These outputs synthesize proposals from group brainstorming and discussion and do not necessarily reflect the views of individual participants or organizations.

“These outputs synthesize proposals from group brainstorming and discussion and do not necessarily reflect the views of individual participants or organizations.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The same breakout groups remained together across the three sessions to support continuity and cumulative sensemaking.

“The same breakout groups remained together across the three sessions to support continuity and cumulative sensemaking.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Ongoing debates around 'responsible' open source hinge not only on practices of release and transparency, but on who gets to decide what counts as responsible, whose values those definitions reflect, and whose interests they serve.

“Ongoing debates around”

Classified as probably not a substantial claim.

The workshop outputs form a layered agenda anchored by a co-created vision, with actions and a research roadmap.

“Together, these outputs form a layered agenda anchored by a foundational north star (co-created vision), interventions to get there (actions), and extended through future-oriented inquiry (research roadmap).”

Probably not a substantial claim: may not pass S3 (load-bearing).

An established body of work has long called for envisioning alternative technological futures, particularly those grounded in diverse social and cultural perspectives.

“An established body of work has long called for envisioning alternative technological futures, particularly those grounded in diverse social and cultural perspectives [20, 23, 27, 28].”

• cites [20] [20] Costanza-Chock, S. Design Justice: Community-Led Practices to Build the Worlds We Need . MIT Press, Cambridge, MA, USA, 2020. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [23] [23] Dunne, A., and Raby, F. Speculative Everything: Design, Fiction, and Social Dreaming . MIT Press, Cambridge, MA, USA, 2013. - sourceNot checked — no full text of the source
• cites [27] [27] Haraway, D. Situated knowledges: The science question in feminism and the privilege of partial perspective. Feminist Studies 14 , 3 (1988), 575-599. - sourceNot checked — no full text of the source
• cites [28] [28] Harding, S. G. Whose Science? Whose Knowledge? Thinking from Women's Lives . Cornell University Press, Ithaca, NY, USA, 1991. -Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The workshop had 75 registered participants, drawn from 88 invitees and 12 nominations.

“Seventy-five participants registered for the workshop, from a pool of 88 individuals on the invitation list and 12 people identified through the nomination form.”

May combine two claims.

Many harms stem less from openness than from availability and capability, with widely available proprietary models also enabling misuse at scale.

“recent work emphasizes that many harms stem less from openness than from availability and capability, with widely available proprietary models also enabling misuse at scale [16].”

• cites [16] [16] Casper, S., O'Brien, K., Longpre, S., Seger, E., Klyman, K., Bommasani, R., Nrusimha, A., Shumailov, I., Mindermann, S., Basart, S., Rudzicz, F., Pelrine, K., Ghosh, A., Strait, A., Kirk, R., Hendrycks, D., Henderson, P., Kolter, Z., Irving, G., Gal, Y., Bengio, Y., and Hadfield-Menell, D. Open technical problems in open-weight ai model risk management. Social Science Research Network (2025). - sourceNot checked — no full text of the source

May combine two claims.

Opening up deliberation supports collective problem framing and prevents premature closure around singular solutions.

“opening up' deliberation to support collective problem framing and prevent premature closure around singular solutions [52]”

• cites [52] [52] Stirling, A. 'opening up' and 'closing down': Power, participation, and pluralism in the social appraisal of technology. Science, Technology, & Human Values 33 , 2 (2008), 262-294. Published online in 2007. - sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

What counts as 'open', who decides where systems fall along the openness spectrum, and who has the capacity to meaningfully engage with them shape which futures become imaginable, legitimate, and actionable.

“What counts as 'open', who decides where systems fall along the openness spectrum, and who has the capacity to meaningfully engage with them shape which futures become imaginable, legitimate, and actionable.”

May combine two claims.

Historically, open source has helped diffuse innovation, challenge incumbents, and enable viable alternatives in otherwise concentrated markets.

“Historically, open source has helped diffuse innovation, challenge incumbents, and enable viable alternatives in otherwise concentrated markets [61].”

• cites [61] [61] Weber, S. The Success of Open Source . Harvard University Press, Cambridge, MA, USA, 2004. - sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Interpretation was informed by the researcher's background and iterative discussions with co-authors supported reflexivity and rigor.

“Interpretation was informed by this researcher's scholarly background on open source, responsible AI, and sociotechnical questions of power and fairness, while iterative discussions with co-authors, who also participated in the workshop, supported reflexivity and rigor in interpretation.”

May combine two claims.

The four tensions reveal conflicting interpretations of openness regarding its purpose (as an end or a means), its scope (expansion versus meaningful access), and its operation (mandatory versus conditional, sufficient versus dependent on governance and use).

“revealing conflicting interpretations of openness regarding its purpose (as an end or a means), its scope (expansion versus meaningful access), and its operation (mandatory versus conditional, sufficient versus dependent on governance and use).”

Probably not a substantial claim: may not pass S2 (checkable).

Definitions of openness are not merely technical, but sites of agenda-setting power.

“Taken together, these dynamics illustrate that definitions of openness are not merely technical, but sites of agenda-setting power.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Open AI systems offer practical advantages by enabling reuse, adaptation, and fine-tuning of existing tools that can help mitigate the consequences of persistent asymmetries in access to compute and infrastructure that remains consolidated in the Global North.

“Open AI systems offer practical advantages by enabling reuse, adaptation, and fine-tuning of existing tools that can help mitigate the consequences of persistent asymmetries in access to compute and infrastructure that remains consolidated in the Global North.”

Probably not a substantial claim: may not pass S2 (checkable).

Definitions and practices of openness and responsibility in AI are not neutral technical classifications, but sites of agenda-setting power.

“definitions and practices of openness and responsibility in AI are not neutral technical classifications, but sites of agenda-setting power.”

May combine two claims.

Even when code or models are open, limited access to compute constrains participation and concentrates influence among a small set of actors.

“Even when code or models are open, limited access to compute constrains participation and concentrates influence among a small set of actors.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Participants worked in pre-assigned breakout groups of seven to nine people to ensure diversity across affiliation, sector, and demographic background.

“Participants worked in pre-assigned breakout groups of seven to nine people to ensure diversity across affiliation, sector, and demographic background.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The participant group included global representation from Brazil, France, Germany, Mexico, Nigeria, and the United Kingdom, though most were based in the US.

“While most participants were based in the US, there was global representation from people based in Brazil, France, Germany, Mexico, Nigeria, and the United Kingdom.”

May combine two claims.

Rather than ignoring tensions or treating divergence as a barrier to progress, plurality can be a productive condition: surfacing underlying values, inviting deliberation, and informing paths forward.

“Rather than ignoring tensions or treating divergence as a barrier to progress, plurality can be a productive condition: surfacing underlying values, inviting deliberation, and informing paths forward.”

Classified as probably not a substantial claim.

Methodologically, this paper demonstrates the role participatory futures and design approaches can play in AI governance and agenda setting.

“Methodologically, this paper demonstrates the role participatory futures and design approaches can play in AI governance and agenda setting.”

May combine two claims.

Debates around open source AI cannot be reduced to a simple corporate versus non-corporate divide, but instead reflect deeper differences about values and purposes for openness to serve.

“Tensions also reveal that debates around open source AI cannot be reduced to a simple corporate versus non-corporate divide, but instead reflect deeper differences about values and purposes for openness to serve.”

Probably not a substantial claim: may not pass S2 (checkable).

Disagreements about aspects such as whether openness should be compulsory or conditional are not merely semantic, but consequential for governance and accountability.

“Disagreements about aspects such as whether openness should be compulsory or conditional are not merely semantic, but consequential for governance and accountability.”

Probably not a substantial claim: may not pass S2 (checkable).

Findings reveal that confronting concentration of agenda-setting power in open AI requires ongoing, inclusive governance processes that can hold plural values, negotiate tradeoffs, and support multiple configurations of openness across contexts.

“More broadly, findings reveal that confronting concentration of agenda-setting power in open AI requires ongoing, inclusive governance processes that can hold plural values; negotiating tradeoffs; and supporting multiple configurations of openness across contexts.”

May combine two claims.

The ideas about openness being conditioned on participation and about openness as a primary mechanism for collective governance are not conceptually incompatible, but they reflect different assumptions about whether openness alone is sufficient to redistribute power.

“While these ideas are not conceptually incompatible, they reflect different assumptions about whether openness alone is sufficient to redistribute power.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Rather than seeking resolution, the workshop sought to treat differences in technical expertise as sites of negotiation and embrace plurality.

“Rather than seeking resolution, the workshop sought to treat these differences as sites of negotiation and embrace plurality.”

Probably not a substantial claim: may not pass S2 (checkable).

Tensions require ongoing discussion, negotiation, contestation, and governance to inform how openness is pursued, supported, or constrained.

“Importantly, tensions require ongoing discussion, negotiation, contestation, and governance to inform how openness is pursued, supported, or constrained.”

May combine two claims.

Our contribution is empirical, showing that plurality need not prevent shared understanding, but can produce shared orientations that can guide action and research without premature closure.

“Our contribution is empirical, showing that plurality need not prevent shared understanding, but can produce shared orientations that can guide action and research without premature closure.”

The quote only partly states this claim.

A central challenge is not in identifying a single path to responsible openness in AI, but making plurality explicit and governable.

“This inherent flexibility reinforces their power to define what counts as”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

This work argues for AI governance approaches that embrace plurality, make tensions visible, and clarify how decisions about openness are made, contested, and managed.

“This work argues for AI governance approaches that embrace plurality, make tensions visible, and clarify how decisions about openness are made, contested, and managed.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Findings may be influenced by selection bias, though the authors sought to mitigate this through snowball sampling and recommendation approach.

“Further, findings may be influenced by selection bias, though we sought to mitigate this through the snowball sampling and recommendation approach.”

Classified as probably not a substantial claim. May combine two claims.

Participatory design methods surface collective and situated insights that individual-level methods cannot replicate, but they are not designed for statistical generalizability.

“Additionally, while participatory design methods surface collective and situated insights that individual-level methods cannot replicate, they are not designed for statistical generalizability.”

May combine two claims.

A key limitation is that workshop participants were primarily US-based and did not fully reflect the breadth of the global open source and AI ethics communities.

“While our workshop included global representation, a key limitation is that workshop participants were primarily US-based and did not fully reflect the breadth of the global open source and AI ethics communities.”

May combine two claims.

The plural imaginaries reflect different situated knowledges and informed the action pathways and research priorities.

“These plural imaginaries reflect different situated knowledges and informed the action pathways and research priorities.”