I spoke with Christian Kleinerman, EVP of Product at Snowflake, at a recent industry event in San Francisco. We discussed open source’s key role in the development of artificial intelligence and the importance of an agnostic approach to hosting AI models. Below are the key points and select quotes from our conversation.
Key Points: AI and Innovation
Checks and balances: Open source keeps proprietary AI vendors in check and drives innovation.
Model-agnostic: Snowflake adopts a model-agnostic approach, offering both open source and proprietary AI models to customers while providing security and governance guarantees.
Agentic AI: Open source plays an important role in the development of agent-based AI workflows that aim to automate more complex tasks.
Responsible AI: Responsible AI practices, including transparency around model training and data usage, are becoming increasingly important, driven in part by the focus on privacy and ethics.
Interoperability and portability: The future of data and AI architectures will emphasize interoperability and portability, enabled by open data formats like Apache Iceberg, as well as open source catalog solutions like Apache Polaris.
Select Quotes: “AI-Driven Democratization”
Along with discussing how open source drives innovation, Kleinerman also detailed open source’s role in AI, the importance of being model-agnostic, and how AI democratizes data.
The Role of Open Source in AI Innovation
“Think about the role of open source in AI and innovation — I think it’s fascinating what’s going on. The reality is, almost once a week or every other week, we all wake up to a different new reality, a new innovation, and a different leader in the space. I think it’s very important because it keeps all the proprietary vendors in check and it puts pressure on all of them.
“Much was said about when DeepSeek came out, but I would say even Llama in the original days was as impactful and shifting as many other proprietary launches and announcements.
“The most important thing is that this is an area that is fast-moving, fast-evolving, fast-innovating, and as long as it doesn’t become a world where two or three companies control these models, that’s good for all of us — beyond just the data industry. Open source is playing a very important role there.”
Snowflake’s Model-Agnostic Strategy and Enterprise Trust
“At Snowflake, we’re model agnostic — hosting both open source as well as proprietary models. When we tell our customers, ‘you can use model X, Y, Z in Snowflake within Snowflake’s security perimeter,’ we provide the same guarantees across all models: we will not train on your data, we will not leverage the outputs of your questions for other customers’ benefit. That is what really matters, and that’s what organizations and enterprises truly care about.
The distinction often made between open source and proprietary models actually misses the point — it’s not about whether the model is open or closed, it’s about who is hosting it and what their guarantees are.
“For example, DeepSeek is an open source model. We took the weights of that model and can offer all the Snowflake security guarantees. But if you access an endpoint hosted elsewhere, say in China, then yes, there may be other risks. The key is customer trust and control — that’s what we’re focused on.”
The Future of Open Data and AI-Driven Democratization
“The future of data is open and interoperable — that’s very clear. Anyone managing a medium-to-large data estate wants to mix and match technologies, avoid lock-in, and keep their data in formats and systems they control. That’s why open table formats like Apache Iceberg are so important, and why Snowflake is deeply committed to enabling that world.
“But there’s a truth that isn’t said enough: today, there are still trade-offs. If I choose interoperable formats, many engines don’t yet agree on how to handle permissions, column masking, row filtering — basic governance. That’s the next frontier.
“As for AI, it’s moving at breakneck speed — what’s possible changes every three months. But what I can say confidently is that AI will bring the next level of democratization to data. Just like business intelligence tools empowered people to get insights without needing a technical team, AI will enable people to simply ask questions in natural language and get answers — even if they don’t know how to use Tableau or Power BI. It’s transformative for access, inclusion, and productivity.”
