OpenAI goes open source: Nearform experts share how enterprises can benefit - Part 1OpenAI goes open source: Nearform experts share how enterprises can benefit - Part 1OpenAI goes open source: Nearform experts share how enterprises can benefit - Part 1

This week, OpenAI launched two of its AI models to the public, releasing gpt-oss as a free download. A private company spending billions of dollars in R&D to then release its product to the public for free may not be an intuitive decision, but additional context on the nature of the open source community could shed light on how this benefits OpenAI - and developers everywhere.

What we know so far from this announcement is that it means no more API costs for developers - rather, it’s a one-time hardware investment that enables infinite usage. Data has privacy by default - a critical advancement for regulated industries with sensitive data and/or compliance requirements (think healthcare, finance, legal, and government). It works offline - that’s right, no internet required to access, which can potentially open the door to new markets and audiences. And the model itself can be modified - not just the prompts. Organizations can rethink their digital business models, selling (one-time) apps instead of subscriptions, potentially at premium prices. In short - there is a wide number of options and choice points available today that weren’t available earlier this week.

What does this release mean for developers, and for companies ranging from start-ups to enterprises? Nearform has been a global leader in open source contributions since its founding over a decade ago, and today two Nearform experts weigh in on the ever-changing landscape of open source in AI: Sarah Michelle Egan, Technical Director based in the US, and Colm Harte, Head of Delivery for Europe, based in Ireland.

We had a lengthy discussion on all things open source, AI, and Open AI - available in two parts. Here’s part 1: Must-know implications of Open AI’s recent launch into open source.

Sarah Egan: People will see an uptick in adoption across domains because this is allowing different industries to become involved with AI that weren't before.

I think it'll lead to more people feeling like they can adopt AI in their companies and in their workflows because their teams will be excited about it and feel like they have more ownership over the tooling.

When you're considering a model for an organization, especially a proprietary model, there's a lot of red tape - getting licensing fees and going through legal procurement and all those things. And that would still happen with an open source tool like this, but the guardrails are a lot less. You can start playing around with it on smaller teams, on teams that maybe don't have the financial backing to go and and use a proprietary model that's more expensive. And so even if it's in small pieces, we'll see a lot more innovation from those domains and industries.

That’s why I could see [the Open AI release] leading to an uptick in adoption, especially in regulated industries that maybe were hesitant to adopt, when with proprietary models they felt they can analyze the risk a little bit better. Now they can train the model on their own data if they want. And so that extra transparency, I think will be important for some clients. When a library or package has really great involvement like that, it's easy for companies to sign on because there's like a lot of backing. And with OpenAI releasing this, it's going to have really great backing from them and then the community as a whole.

One, there's more transparency with the model, and that's a great thing. But then there's the downside to having it be open source: the proprietary models have a lot of monitoring and governance, and releasing, that companies do themselves. With an open source model, now that responsibility will be passed on to the companies that are using it. So that's something that they'll have to keep in mind of like building into their pipeline of of tooling and consider building some of their own controls to track like latency and drift and just how the model evolves, because those are things that OpenAI would be doing on their own for their proprietary models. But I think that the benefits still outweigh any cons or any downsides to it.

And two, scalability definitely is something that these companies will have to consider if the open source model can handle their production workload and workflows, because it might not be quite the same as a proprietary model. That would be something that they have to do some extra discovery on maybe at the beginning.

Security and compliance is another huge consideration, obviously even more so with AI. Outside of how the models are actually functioning and performing, security is the biggest priority when you're considering choosing a model. So designing for that from the beginning is going to be really important.

Colm Harte: It gives enterprises additional options that they didn't previously have. You have full control if you have the open source model and you manage it yourself, you know you can rely on, or you're in control of how your own data is then managed. You're not dependent on the third party in terms of what they may or may not do with your data. So I think from that perspective, what it's giving enterprises is, capability that they can now leverage more comfortably.

So you might have had enterprises that said, ‘look, we can't really go on the LLM journey because our data is too sensitive. We can't trust it. We know it goes against all our data access policies, so we can't take advantage of these solutions.’ But this now gives them a way to take advantage of the whole GenAI wave, the capabilities, and improve their own products as a result of that.

It could become much more cost-effective if you've got a hugely high volume and then actually there's an initial investment to get it running in your own enterprise infrastructure, but over time that's going to pay back because you're not paying licensing fees to OpenAI. Yeah, it just gives them more control, and I think for some enterprises, that will be a key consideration for them.

SE: And I think when any tool is open sourced, it feeds into this sense of: “we're all in technology, in this community”. And when something is opened up for us to comment on or play around with, it shows commitment to that community, putting it back into the hands of the developers and the technologists who are using these on the ground and and can allow us to make some decisions (or at least like feel like we're a part of it), and that's really exciting.

Open source is very near and dear to Nearform’s heart, and mine in particular. I've been an open source maintainer - it's the best. It's all about community and contributing. And when I'm suggesting tools to clients, especially open source ones, the considerations we make are, you know, how many people are using it? What's the maintenance like? What’s the community involvement like, with people updating documentation and issues and things?

CH: There's a long history of proven value in open sourcing solutions. But what they’re really giving is more optionality to enterprise and customers and consumers, as to how they want to engage with these types of systems.

Some organizations may think, “I'm quite happy with the proprietary model. It does a lot of heavy lifting, and it’s effective from a cost perspective. So I'm not going to take on the burden of running things on my own.” But as they grow and expand, and learn about these systems as their own usage increases, they might decide at a later stage that it makes a lot more sense now to run it directly using open source. So having these versions available just gives them that capability.

SE: Not just for the open source AI model, but also for any open source tool - when you remove vendor lock-in or paid subscriptions, it levels the playing field. That means smaller teams, or in-house teams, can embrace really advanced capabilities without the financial burden of licensing fees. There may be some upfront costs for infrastructure, but those investments are easier to see over time. I think we'll see domain-specific innovation across industries that previously were hesitant or unable to adopt AI.

And there's a huge market of internet users with subpar network access. I think we always think about our own systems and our own devices, but there's millions of people out there who are just on mobile devices and don't have the internet connectivity that we might be used to. And those people can be reached with open source models because of their offline capabilities. And that'll be a big advancement for the use of certain tools, or for certain companies looking to expand into markets they couldn't before.

SE: I love this part. I think contribution is the heart of open source and the community. And obviously at Nearform we have a really profound history of open source engagement. So using the tools is the first step to making them better.

It could be improving documentation, like documenting issues that you run into with test cases. You don't have to be a high level AI developer to contribute to AI open source. I think it's really valuable to invest in these tools to solve shared problems. And companies can actually give back while they're solving their own goals. And those contributions help your own company, but also everyone else using it - it’s truly an exciting time.

CH: I think what you'll see is that companies will innovate, and they will come up with tooling and modules components that will help them leverage these open source systems. Open sourcing things as they create them is a great way to contribute back to the open source community.

Coming soon Part 2: Recommendations for enterprise adoption of open source in AI.