Future of AI: Is Open-source the key to ethical and transparent development?

The landscape of artificial intelligence is at a crossroads, balancing between proprietary and open-source development. While major corporations like Google, OpenAI, and Microsoft continue to advance cutting-edge AI through closed models, an equally powerful movement advocates for open-source AI as a more transparent and accessible alternative. The question remains: is open-source AI the future?

A recent study titled Is Open Source the Future of AI? A Data-Driven Approach by Domen Vake, Bogdan Šinik, Jernej Vicič, and Aleksandar Tošic, published in Applied Sciences (2025, 15, 2790), explores this debate using empirical evidence.

The open-source advantage: Transparency and innovation

One of the core arguments in favor of open-source AI is its ability to foster transparency and innovation. Unlike proprietary AI models, which often operate in a "black box" manner, open-source AI allows for community-driven improvements, independent audits, and wider accessibility. This study highlights how platforms like Hugging Face have become central hubs for AI collaboration, enabling researchers and developers to contribute enhancements to large language models (LLMs).

The research found that open-source modifications to existing LLMs often result in efficiency gains without sacrificing performance. For instance, models such as Llama and Mistral have benefited from community-driven enhancements that optimize their capabilities while remaining cost-effective. Additionally, open-source models serve as educational tools, allowing universities and independent researchers to study and refine AI methodologies without the barriers of proprietary restrictions.

Proprietary vs. open-source AI: The performance gap

While open-source AI offers accessibility and collaboration, proprietary AI models currently maintain a lead in performance. The study references research from Epoch AI, which estimates that open-source models lag behind proprietary AI by roughly 15 months. However, this gap is steadily narrowing as open models continue to improve through collective efforts.

The research underscores how corporate-backed AI models have the advantage of immense computing power, private datasets, and dedicated teams of researchers. Open-source models, on the other hand, must rely on publicly available datasets and distributed computing resources. Despite this, community-led improvements have significantly reduced performance disparities, making open AI models viable for various applications. Additionally, hybrid approaches - where companies release partially open models with restricted licenses - are gaining traction as a middle ground.

Risks and challenges of Open AI

Despite its benefits, open-source AI is not without challenges. One major concern is the potential for misuse. Open AI models, once publicly available, can be leveraged for unethical purposes, including deepfake creation, automated disinformation campaigns, and malicious applications. The study acknowledges these risks and discusses potential safeguards, such as licensing agreements that impose ethical usage restrictions.

Another challenge is the lack of financial incentives for open-source development. Unlike proprietary AI, which generates revenue through subscriptions and enterprise licenses, open models often rely on voluntary contributions and grants. Sustaining long-term innovation in open-source AI requires new business models that align financial sustainability with accessibility.

Future of AI: A hybrid model?

The study concludes that the future of AI may not be exclusively open or closed but rather a hybrid approach. Many organizations are adopting a mixed strategy, releasing some aspects of their models while keeping sensitive components proprietary. For example, models like Mistral and Llama provide open weights but restrict access to training data, allowing for community collaboration while maintaining some control over proprietary elements.

As AI development continues to evolve, policymakers and industry leaders must find a balance between openness and regulation. Open AI contributes to global AI democratization, but responsible governance is crucial to mitigating risks. Ultimately, the trajectory of AI development will be shaped by technological advancements, ethical considerations, and the interplay between corporate innovation and community-driven progress.