Privacy-first AI: Can decentralised models beat ChatGPT?

In a recent panel discussion titled "Privacy-first AI: Can decentralised models beat ChatGPT?", during our deAI Summit 2025 with Pundi AI, AI founders and leaders shared insights on scaling AI solutions, with a focus on the comparison between decentralised AI tools and private, secure models.

Subscribe to our YouTube channel for the latest AI content and insights. Connect with us on LinkedIn, X, Telegram, WhatsApp, TGB website.

Panelists:

The discussion delved into critical questions surrounding data privacy, decentralisation, regulatory compliance, and the future landscape of AI models. The insightful conversation featured prominent voices in the AI space:

• Jackie Tan (CEO, Tiger and Wolf - a boutique AI consultancy)

• Calvin Tan (CTO, Pints AI - an AI platform for financial and insurance)

• Tina Chopra (CEO, Addlly AI - creating AI agents for marketing)

• Mckenzie Mathieson (Regional Head, Airia AI - AI orchestration platform)

• [Moderator] Shivang Gupta (CEO, The Generative Beings)

Key Insights:

• The Rise of AI in Southeast Asia: The presence of international panelists and the establishment of AI companies in Singapore highlight the region's growing importance as a hub for AI innovation.

• Balancing Decentralisation and Centralisation: panelists emphasised that enterprises are seeking a balance between the transparency of decentralised models and the control of centralised ones. The choice of model ultimately depends on the specific use case and the need for data sovereignty.

• The Importance of Production-Ready AI: panel pointed out that many AI tools focus on prototyping, neglecting the crucial step of creating production-ready and secure AI solutions. Area aims to fill this gap by providing a secure foundation for enterprises to build and deploy AI agents.

• AI-Powered Content Creation at Scale: panelists addressed the challenge of creating on-brand content at scale. They explained that while custom GPTs work for smaller enterprises, larger companies require more secure, brand-specific solutions.

• The Future is a Hybrid Approach: The panelists generally agreed that the future of AI is not a zero-sum game between decentralised and centralised models. Instead, a hybrid approach, where different models are used for different purposes, is the most likely path forward. The ultimate goal is to solve the client's problem, regardless of the underlying technology.All sessions are taught by industry experts and pioneering founders

Q&A with speakers:

Q: Are enterprises demanding data segregation, transparency, and decentralisation or just chasing centralised models?

A (Mckenzie): Enterprises want it all - data segregation, transparency, and decentralisation for regulatory comfort and control. Decentralised models help on transparency and sovereignty, but most organisations seek a blend, not a binary pick. The main shift is focus: buyers care more about building agents and automation over the underlying model’s dataset. They want the right building blocks for their use case.

Q: Does a decentralised approach help brands maintain consistency when creating content at scale?

A (Tina): Big brands prioritise security and integrated workflows. Their needs depend on AI maturity, but most want productivity, AI efficiency, without high engineering overhead. Decentralised approaches can help with data protection and multi-model choices, but constant content rewrites mean controlling the narrative is crucial. As content discovery shifts toward conversational search (GEO), brands need to optimise for these new interfaces.

Q: How do decentralised models advance regulatory compliance and auditability versus ChatGPT?

A (Calvin): ChatGPT and other centralised platforms present data custody risks, concentrating too much control. Decentralisation, especially via Web3 tech, allows for self-regulation—no one “owns” or withholds data. Corporates are eyeing secure and traceable decentralised chains. While traditional Web2 privacy uses on-premise, decentralised networks deploy encryption, so nothing is hostage.

Q: Can decentralised models give creators a better playground compared to closed markets - especially for productivity vs. creativity?

A (Jackie): For creatives, vanilla LLMs are frustrating because output attribution is missing - creators feel “erased.” Decentralised AI offers a future where creators own and track their work, assigning real value back to the source. Though reminiscent of early NFT promises (whose results were mixed), decentralised models still aim to deliver true attribution and fair monetisation.

Q: Should creators build on decentralised/open source models for IP protection or stick to private clouds?

A (Jackie): The outcome of high-profile lawsuits will shape the field for years. If the courts set strong precedents, decentralised models could favor creators and rights holders. Lessons from NFTs: real benefits often went to a small subset. Until the dust settles, creators must weigh protection and monetisation between open ecosystems and private clouds.

Q: What’s next - will open-source win, or will we see a hybrid future with big enterprise models?

A (Calvin): It’s unpredictable. Open-source is invaluable for niche, specialised domain - think healthcare or research. Just as Web3 broke finance barriers, open AI models could break labor bottlenecks.

A (Tina): It’s a moving target. We work across open and closed models by use case. Clients care about cost, security, context, and whether it runs on-premise or SaaS.

A (Mckenzie): Expect a hybrid world - just like how the cloud replaced on-premise, then both coexisted. Decentralisation adds transparency and sovereignty, but centralised models offer enterprise scale. Choice matters most.

Moderator’s Conclusion:

Both centralised and decentralised models will coexist - each solves different problems for different people. The winning approach? Map the tech stack to the client’s use case: solve for data control and transparency, or scale and tooling, as needed. Teams should aim for solutions, not dogmas.

Disclaimer: This summary has been edited and generated with the assistance of AI tools. All interpretations, paraphrasing, and opinions presented are for informational purposes only and should not be considered legally binding or construed as official positions of any panelist, The Generative Beings, or event organizers. Neither the panelists nor TGB nor the organisers accept any responsibility or liability for the content herein.

Become a better being, with generative AI.

— Team Generative Beings