Introduction

Artificial intelligence (AI) has transcended its status as a specialized research field to become a fundamental driver of technological and social innovation. A particularly relevant subset of this discipline is open-source AI (OSAI), whose proliferation is redefining paradigms of development, access, and technological governance (Hermansen & Osborne, 2025). Open source code allows for public examination, modification, and distribution of AI systems, contrasting with proprietary models that restrict access to their internal components.

Definition and context of open source AI

Open-source AI refers to artificial intelligence systems, including models, algorithms, and datasets, whose source code is publicly accessible and can be used, studied, modified, and distributed by anyone for any purpose (Zacchiroli, 2025). This approach is rooted in the principles of open-source software, promoting transparency and collaboration. It involves the availability of pre-trained models, network architectures, libraries, and tools that facilitate the construction of AI solutions (Carter & Hermansen, 2026). Implementing specific licenses is crucial to balancing openness with commercial viability and intellectual property protection, addressing unique characteristics such as data dependencies and model reproducibility (Adebiyi & Adeusi, 2025a).

Growing importance in the current technological landscape

The widespread adoption of open-source AI is due to multiple factors. Organizations of all sizes use some form of open source in their AI technology stacks (Hermansen & Osborne, 2025). This phenomenon lowers barriers to entry for developers and companies, democratizing access to advanced AI capabilities. The open-source community has proven to be a catalyst for innovation, enabling faster iteration and continuous improvement of models (Vake et al., 2025). The tension between openness and intellectual property protection is a recurring theme, as seen in the evolution of organizations moving from a purely open model to more controlled licensing approaches (L, 2025)(L, 2025a).

Purpose and scope of the article

This article comprehensively examines open-source AI, analyzing its evolution, underlying principles, and multidimensional implications. Key themes will be considered, including system transparency and explainability, economic and labor impacts, ethical and intellectual property considerations, and current technological trends. Finally, a synthesis of the benefits and challenges will be presented, along with recommendations and perspectives for the future of open AI.

Overview and evolution of open source AI

Open-source AI represents a convergence of free software development principles and the transformative capabilities of artificial intelligence. Its trajectory has been shaped by the interaction between the community, academia, and industry.

Origins and historical development

The foundations of open-source AI lie in the philosophy of free software, which promoted the freedom to use, study, modify, and distribute computer programs. With the emergence of modern AI in the 21st century, especially with advances in deep learning, the need for resource sharing became evident. Platforms like Hugging Face have greatly facilitated collaboration and the sharing of deep learning models, particularly Large Language Models (LLMs) (Quillivic & Mesmoudi, 2024)(Azarm et al., 2024). From 2012 to 2024, a notable increase in the development of open-source AI models was observed, with substantial growth after 2020, attributable to the evolution of transformational models (Azarm et al., 2024)(Azarm et al., 2025).

Fundamental principles: transparency, collaboration and access

Open-source AI is governed by transparency, which allows users to inspect the inner workings of models, facilitating the identification of biases or vulnerabilities (Potter et al., n.d.). Collaboration is manifested in communities of developers who contribute to improving and extending existing systems, which can significantly enhance model performance (Vake et al., 2025). Universal access to these technologies reduces barriers to entry, allowing a broader range of researchers and small businesses to participate in AI development, thus promoting technological democratization (Hermansen & Osborne, 2025).

Differences between open-source AI and proprietary models

The main distinction lies in control and accessibility. Proprietary models, developed and maintained by private entities, restrict access to source code and training data, often under strict commercial licenses. This can limit external verification, adaptability, and innovation by third parties (Vake et al., 2025b). In contrast, open-source AI, as the name suggests, liberates these elements, fostering an ecosystem where the community can contribute to improvement and customization. However, this openness also presents challenges, including potential misuse, financial disincentives for private research, and intellectual property concerns (Vake et al., 2025b). The current trend shows a spectrum between complete openness and exclusive ownership, with hybrid approaches seeking a balance (Vake et al., 2025b)(Vake et al., 2025).

Key topics in open-source AI

The evolution of open-source AI has led to multifaceted discussions addressing technical, social, economic, and ethical aspects. These issues define the framework for its responsible development and application.

Transparency and explainability in AI systems

The transparency of AI systems refers to the ability to understand how they work, while explainability relates to the ability to communicate their behavior in a way that is intelligible to humans (Shukla, 2024). In the context of open-source AI, the availability of source code inherently facilitates transparency. However, the complexity of many models, especially LLMs, can make them difficult to interpret, even with access to the code (Thelwall, 2025).

Understanding challenges for end users

Despite open-source code, the intricate architecture of advanced AI models, such as deep neural networks, often makes their inner workings a “black box,” even for experts. For end users without deep technical knowledge, understanding the decisions of an AI system can be particularly difficult (Mark & Kobsa, 2005). This raises concerns about trust, accountability, and the ability to challenge biased or erroneous results. The lack of standardized documentation and clear protocols for training data exacerbates this problem (Azarm et al., 2024)(Azarm et al., 2025).

Recommendations for effective AI communication

To address the challenges of understanding AI, it is crucial to develop methods that improve its communicability. This includes creating more intuitive interfaces and providing clear and concise explanations of how predictions are obtained. Research suggests that transparent, self-correcting systems are fundamental to group performance (Mark & Kobsa, 2005). Furthermore, ethical frameworks that prioritize transparency and accountability from the design phase are recommended (Shukla, 2024). Standardizing documentation and information on model training would also contribute significantly (Azarm et al., 2024; Azarm et al., 2025).

Economic and workforce impact

Open-source AI exerts considerable influence on the global economy and labor market dynamics. Its benefits range from process optimization to the creation of new opportunities, though not without generating debates about job restructuring.

Business adoption and innovation acceleration

A significant percentage of organizations (89%) integrate some form of open source into their AI systems, and nearly two-thirds (63%) employ open models (Hermansen & Osborne, 2025). This adoption is based on the economic benefits offered by open source AI, which is considered a cost-effective alternative to proprietary solutions. It facilitates collaboration and accelerates innovation by allowing multiple stakeholders to contribute to and benefit from a common set of tools and models (Hermansen & Osborne, 2025).

Effects on productivity, costs, and collaboration

Open-source AI contributes to increased productivity. For example, in the manufacturing sector, investment in research and development, often associated with AI, shows a positive correlation with labor productivity (Parmar & Stephen, 2024). It also reduces development and implementation costs, as companies do not need to invest from scratch in creating fundamental algorithms or models. The collaborative nature of open source fosters knowledge sharing and co-creation, which can lead to more robust and adaptable solutions (Hermansen & Osborne, 2025).

Sectoral implications: health, agriculture, energy and more

Open-source AI has specific implications across various sectors (Hermansen & Osborne, 2025). In healthcare, it can improve diagnoses and treatments (Gerrard et al., 2023)(Gerrard et al., 2023a), while in agriculture, it can optimize production and resource management (Bernard de Raymond et al., 2021). In the energy sector, it can contribute to more efficient infrastructure management. The ability to adapt and customize open models for specific sectoral needs accelerates digital transformation in multiple domains, such as the pharmaceutical industry, where its impact on innovation capacity has been investigated (Federle, 2024)(Federle, 2024a).

Ethics, intellectual property and governance

The rise of open-source AI intensifies the debate on how to balance innovation with responsibility, the protection of rights, and effective governance. These aspects are fundamental to the sustainable development of technology.

Legal implications and regulatory challenges

The regulation of AI, especially open-source AI, presents significant challenges. Existing laws often fail to adequately address the complexities of AI-generated content, authorship, and accountability (Khadka, 2025). Initiatives such as the European Union’s AI Act and the US Blueprint for an AI Bill of Rights seek to establish regulatory frameworks (Sanwal, 2023)(Manish, 2024). However, the variety of interpretations across sectors and regions hinders the creation of globally accepted ethical standards (Burdzhiev et al., 2025). A regulatory framework that ensures consistency and fairness across all AI-related instances is essential (Sharma, 2025).

Intellectual property in the development and use of open AI

Intellectual property (IP) in open-source AI is a complex area. While source code may be open, training data and resulting models may have different IP regimes. This raises questions about authorship and ownership of AI-generated results (L, 2025)(L, 2025a). Open-source licenses for AI seek to balance knowledge sharing with the protection of interests, but they must consider the particularities of AI systems, such as data dependency and reproducibility (Adebiyi & Adeusi, 2025)(Adebiyi & Adeusi, 2025a)(Olajumoke Ifeoluwa Adebiyi & Oluwafemi Clement Adeusi, 2025). The decision by some organizations to transition from an open approach to a controlled licensing model, as in the case of OpenAI, reflects the tension between openness and IP protection (L, 2025)(L, 2025a).

Ethical principles: fundamental rights, equity and control

Ethics in AI focuses on ensuring that AI systems respect fundamental rights, promote fairness, and allow for human oversight. Open-source AI, with its inherent transparency, can facilitate auditing and bias detection, contributing to fairer AI (Potter et al., n.d.) (Widder et al., 2022). However, it also faces challenges, such as potential misuse or the spread of misinformation. Ethical frameworks, such as those proposed at the EU level, advocate for principles such as respect for human rights, non-discrimination, quality, safety, transparency, and user control (BÎLBĂ, 2024). Harmonizing ethical codes is crucial for interdisciplinary work and for balancing innovation with public safety and trust (Burdzhiev et al., 2025).

Emerging trends and featured tools

The field of open-source AI is experiencing a dynamic evolution, marked by the emergence of new models and tools that expand its capabilities and applications.

Recent open models and their relevance (LLMs, vision, etc.)

Large Language Models (LLMs) have garnered academic and industrial attention, with models such as Meta’s Llama 3 and Mistral’s 8x7B. These LLMs have demonstrated remarkable performance, although proprietary models like Google’s GPT-4 and Gemini are also prominent (Schrepel & Potts, 2025). The open-source community has contributed significantly to the improvement of these models, achieving efficiencies without compromising performance (Vake et al., 2025). While there has been an increase in the development of open-source models post-2020, primarily in text processing, the audio and image processing domains have grown more slowly (Azarm et al., 2024; 2025). There is growing interest in models such as LLaMa and BARD, while others like Mistral and Claude remain less explored in research (Billah et al., 2025).

Most influential open-source AI tools and platforms

Hugging Face has established itself as a fundamental platform for the open-source AI community, facilitating the exchange and collaboration in the development of deep learning models, especially LLMs (Quillivic & Mesmoudi, 2024). This platform, along with other open-source software initiatives, enables the creation of a “digital commons” for AI, where resources and data can be managed and maintained by the community (Quillivic & Mesmoudi, 2024)(Zacchiroli, 2025a). The ability of these tools to integrate ethical design principles and ensure security is essential to mitigating risks, including the generation of toxic content or the leakage of private information (Biswas & Talukdar, 2023).

Analysis and implications of open-source AI

Open source AI is a complex phenomenon with a range of implications that deserve detailed analysis, both in its advantages and its potential drawbacks.

Benefits: democratization, accelerated innovation, and reduced barriers

Open-source AI promotes technological democratization by making advanced tools and models accessible to a wider audience. This reduces reliance on single vendors and fosters competition (Hermansen & Osborne, 2025). Open collaboration accelerates innovation, allowing enhancements and new features to be developed at a faster pace than would be possible in proprietary environments (Vake et al., 2025). Furthermore, lowering barriers to entry for developers and small businesses encourages experimentation and diversity in AI applications, driving more inclusive growth (Hermansen & Osborne, 2025).

Risks and challenges: security, misuse, and technological fragmentation

Despite its advantages, open-source AI presents risks. Open source code can, in certain scenarios, be exploited for malicious purposes, such as the development of surveillance systems or the generation of disinformation (Vake et al., 2025b). The security of open AI systems is an ongoing concern, especially when integrated into critical infrastructure. Technological fragmentation, where multiple versions of the same model or tool coexist without clear coordination, can hinder interoperability and maintenance. Furthermore, the lack of robust financial incentives to maintain and update projects can compromise their long-term sustainability (Vake et al., 2025b).

Systemic impact on scientific research and knowledge acquisition

Open-source AI has transformed scientific research. It enables greater reproducibility of experiments and facilitates the validation of results by allowing other researchers to examine and test the code and data. This collaboration expands the scope of research, enabling the exploration of new frontiers of knowledge (Potineni, 2025). The ability to leverage and adapt pre-existing models reduces the time and resources needed for new studies, accelerating discovery in fields such as medicine and astronomy (Potineni, 2025).

Balance between openness, protection and sustainability

Striking a balance between full openness, intellectual property protection, and economic sustainability is a central challenge for open-source AI. Licenses must be designed to foster collaboration and innovation while enabling monetization and the continued development of projects (Adebiyi & Adeusi, 2025a). An approach that considers the needs of the community, industry, and regulators is essential for building a robust and beneficial open AI ecosystem in the long term (Kijewski et al., 2025).

Conclusion

Open-source AI represents a transformative force with profound implications for technology, the economy, and society. Its collaborative and transparent development model offers substantial benefits, but also introduces complexities that require careful attention.

Summary of key findings

It has been established that open-source AI, rooted in principles of transparency and collaboration, has experienced significant growth, particularly in post-2020 LLM development (Azarm et al., 2024)(Azarm et al., 2025). Its widespread adoption by organizations underscores its cost-effectiveness and its ability to accelerate innovation across various sectors (Hermansen & Osborne, 2025). However, the inherent complexity of AI models presents challenges to end-user explainability, even with open source (Thelwall, 2025). Intellectual property and regulatory frameworks are evolving to accommodate the specific characteristics of generative and open-source AI (L, 2025)(L, 2025a)(Khadka, 2025).

Recommendations and future perspectives for open AI

To maximize the benefits and mitigate the risks of open-source AI, several courses of action are suggested. Promoting standardization in model documentation and training protocols is crucial to improve transparency and scientific rigor (Azarm et al., 2024)(Azarm et al., 2025). Developing agile regulatory frameworks is essential to address ethical and legal implications, prioritizing non-discrimination, safety, and user control (BÎLBĂ, 2024)(Burdzhiev et al., 2025). Fostering collaborative ecosystems, through platforms like Hugging Face, will continue to be a pillar for innovation and democratization (Quillivic & Mesmoudi, 2024). Finally, ongoing dialogue between academia, industry, policymakers, and civil society is essential to strike a balance between openness, the protection of rights, and the long-term sustainability of open-source AI (Kijewski et al., 2025).

References

Adebiyi, O.I., & Adeusi, O.C. (2025a). Analyzing the impact of open-source licensing models on AI development, commercialization, and knowledge dissemination in technology sectors. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.17199598

Adebiyi, O.I., & Adeusi, O.C. (2025b). Analyzing the impact of open-source licensing models on AI development, commercialization, and knowledge dissemination in technology sectors. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.17199597

Azarm, M., Sangari, E., & Abughoush, K. (2024). Unveiling the Dynamics of Open-Source AI Models: Development Trends, Industry Applications, and Challenges. ScholarSpace (University of Hawaii at Manoa) .

Azarm, M., Sangari, E., & Abughoush, K. (2025). Unveiling the Dynamics of Open-Source AI Models: Development Trends, Industry Applications, and Challenges. In Proceedings of the Annual Hawaii International Conference on System Sciences . Hawaii International Conference on System Sciences. https://doi.org/10.24251/hicss.2025.582

Bernard de Raymond, A., Alpha, A., Ben-Ari, T., Daviron, B., Nesme, T., & Tétart, G. (2021). Systemic risk and food security. Emerging trends and future avenues for research. In Global Food Security (Vol. 29, p. 100547). Elsevier BV. https://doi.org/10.1016/j.gfs.2021.100547

Bhavyateja Potineni. (2025). AI as a research accelerator: Human-AI synergy in scientific discovery and innovation. In World Journal of Advanced Research and Reviews (Vol. 26, Issue 1, pp. 2937–2948). GSC Online Press. https://doi.org/10.30574/wjarr.2025.26.1.1183

BÎLBĂ, II (2024). ETHICAL PRINCIPLES AND AI IN INTELLECTUAL PROPERTY LAW. SHILAP Journal of Lepidopterology .

Billah, M.M., Hamjaya, H.S., Shiralizade, H., Singh, V., & Inam, R. (2025). Large Language Models’ Trustworthiness in the Light of the EU AI Act—A Systematic Mapping Study. In Applied Sciences (Vol. 15, Issue 14, p. 7640). MDPI AG. https://doi.org/10.3390/app15147640

Biswas, A., & Talukdar, W. (2023). Guardrails for trust, safety, and ethical development and deployment of Large Language Models (LLM). In Journal of Science & Technology (Vol. 4, Issue 6, pp. 55–82). Libertatem Media Private Limited. https://doi.org/10.55662/jst.2023.4605

Burdzhiev, S., Burdzhieva, M.-M., Gotsev, L., Syarova, S., & Angelova, V. (2025). A REVIEW ON AI IN CYBERSECURITY: ETHICAL CHALLENGES AND REGULATORY FRAMEWORKS. Environment Technology Resources Proceedings of the International Scientific and Practical Conference , 2 , 285–291. https://doi.org/10.17770/etr2025vol2.8607

Carter, H., & Hermansen, A. (2026). The Value of Open Source AI for the Canadian Economy: A Review of Industry, Academic, and Open Source Evidence . The Linux Foundation. https://doi.org/10.70828/bgqs8505

Federle, M. (2024a). Assessing the impact of open-source AI technology. Evidence from the Indian pharmaceutical industry. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.14416673

Federle, M. (2024b). Assessing the impact of open-source AI technology. Evidence from the Indian pharmaceutical industry. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.14416674

Gerrard, J., Benson, R., Brown, E., & Varughese, C. (2023a). Capturing the benefits of AI in healthcare for Aotearoa New Zealand – Full report [Techreport]. University of Auckland. https://doi.org/10.17608/k6.opmcsa.24814101

Gerrard, J., Benson, R., Brown, E., & Varughese, C. (2023b). Capturing the benefits of AI in healthcare for Aotearoa New Zealand – Full report [Techreport]. https://doi.org/10.17608/k6.opmcsa.24814101.v1

Hermansen, A., & Osborne, C. (2025). The Economic and Workforce Impacts of Open Source AI: Insights from Industry, Academia, and Open Source Research Publications . The Linux Foundation. https://doi.org/10.70828/itvq4899

Khadka, R. (2025). Navigating the Legal Landscape of AI-Created Content: Intellectual Property, Accountability, and Regulation. In International Scientific Journal of Engineering and Management (Vol. 04, Issue 04, pp. 1–7). Edtech Publishers (OPC) Private Limited. https://doi.org/10.55041/isjem03055

Kijewski, S., McBride, C., Ronchi, E., Ferretti, A., & Vayena, E. (2025). Responsible and Strategic Open Source AI Innovation in Europe and Beyond . https://doi.org/10.3929/ethz-c-000787142

L, B. (2025a). Comprehensive Study of Intellectual Property Rights (IPR) in AI Software – Focus on OpenAI and Comparative Analysis [Techreport]. CERN European Organization for Nuclear Research. https://doi.org/10.5281/zenodo.17119582

L, B. (2025b). Comprehensive Study of Intellectual Property Rights (IPR) in AI Software – Focus on OpenAI and Comparative Analysis [Techreport]. CERN European Organization for Nuclear Research. https://doi.org/10.5281/zenodo.17119581

Manish, S. (2024). ETHICAL AND REGULATORY DEVELOPMENTS IN AI GOVERNANCE . https://doi.org/10.5281/zenodo.15461280

Mark, G., & Kobsa, A. (2005). The Effects of Collaboration and System Transparency on CIVE Usage: An Empirical Study and Model. In Presence: Teleoperators and Virtual Environments (Vol. 14, Issue 1, pp. 60–80). MIT Press-Journals. https://doi.org/10.1162/1054746053890279

Olajumoke Ifeoluwa Adebiyi, & Oluwafemi Clement Adeusi. (2025). Analyzing the impact of open-source licensing models on AI development, commercialization, and knowledge dissemination in technology sectors. In International Journal of Science and Research Archive (Vol. 15, Issue 3, pp. 1103–1113). GSC Online Press. https://doi.org/10.30574/ijsra.2025.15.3.1753

Parmar, P.M., & Stephen, J.J. (2024). IMPACT OF AI ON LABOR PRODUCTIVITY IN THE MANUFACTURING SECTOR IN INDIA. In ShodhKosh: Journal of Visual and Performing Arts (Vol. 5, Issue 1). Granthaalayah Publications and Printers. https://doi.org/10.29121/shodhkosh.v5.i1.2024.4413

Potineni, B. (2025). AI as a research accelerator: Human-AI synergy in scientific discovery and innovation. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.17263426

Potter, Y., Potter, M., & Song, D. (nd). “As an AI, I believe AI models should be open source.” https://www.semanticscholar.org/paper/269150091

Quillivic, R., & Mesmoudi, S. (2024). Is open source software culture enough to make AI a common? https://doi.org/10.48550/arxiv.2403.12774

Sanwal, M. (2023). Ethical and Regulatory Developments in AI Governance. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.15331050

Schrepel, T., & Potts, J. (2025). Measuring the openness of AI foundation models: competition and policy implications. In Information & Communications Technology Law (Vol. 34, Issue 3, pp. 279–304). Informa UK Limited. https://doi.org/10.1080/13600834.2025.2461953

Sharma, G. K. (2025). NEUROGAMING, AI, and LEGAL FRAMEWORKS NAVIGATING DATA PRIVACY, INTELLECTUAL PROPERTY, and ETHICAL CHALLENGES. In Gaming Law Review . SAGE Publications. https://doi.org/10.1177/25725300251400742

Shukla, S. (2024). Principles Governing Ethical Development and Deployment of AI. In International Journal of Engineering, Business and Management (Vol. 8, Issue 2, pp. 26–46). AI Publications. https://doi.org/10.22161/ijebm.8.2.5

Thelwall, M. (2025). Research quality evaluation by AI in the era of large language models: advantages, disadvantages, and systemic effects – An opinion paper. In Scientometrics (Vol. 130, Issue 10, pp. 5309–5321). Springer Science and Business Media LLC. https://doi.org/10.1007/s11192-025-05361-8

Vake, D., Šinik, B., Vičič, J., & Tošić, A. (2025a). Is Open Source the Future of AI? A Data-Driven Approach . https://doi.org/10.48550/arxiv.2501.16403

Vake, D., Šinik, B., Vičič, J., & Tošić, A. (2025b). Is Open Source the Future of AI? A Data-Driven Approach. In Applied Sciences (Vol. 15, Issue 5, p. 2790). MDPI AG. https://doi.org/10.3390/app15052790

Widder, D.G., Nafus, D., Dabbish, L., & Herbsleb, J. (2022). Limits and Possibilities for “Ethical AI” in Open Source: A Study of Deepfakes. In 2022 ACM Conference on Fairness Accountability and Transparency (pp. 2035–2046). ACM. https://doi.org/10.1145/3531146.3533779

Zacchiroli, S. (2025a). Free/Open Source Software in the Era of AI Hype. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.17911743

Zacchiroli, S. (2025b). Free/Open Source Software in the Era of AI Hype. Zenodo (CERN European Organization for Nuclear Research) . https://doi.org/10.5281/zenodo.17911744