Meta can call Llama 2 open source all it wants, but that doesn’t mean it is.

Notice The large Llama-2 language model recently released by Meta is not open source.

Yes, I know people from Meta AI proclaim: “Llama 2 [is] the next generation of our open-source LLM available for research and commercial use.” So what? It’s not.

I can say that I’m also a master carpenter, but that won’t change the fact that if I try to drive a nail into the wall with a hammer, I’m going to hit my thumb every time. Or, in this case, Meta is simply open source washing an open but ultimately proprietary LLM.

Meta trots over AI Llama 2 models and invites developers to ride on them


As Erica Brescia, managing director of RedPoint, the open source-friendly venture capital firm, asked, “Can someone please explain to me how Meta and Microsoft can justify calling Llama 2 open source if it doesn’t actually use OSI [Open Source Initiative]- license approved or OSD compliant [Open Source Definition]? Do they intentionally challenge the definition of free software [Open Source Software]?”

I don’t think Meta and its partner Microsoft are intentionally wrong with open source. Their programmers certainly know better, but at the highest level, open source is just marketing jargon.

As OpenUK CEO Amanda Brock put it, it’s “not an OSI-approved license, but a important release of open technology … It’s a step to moving AI from the hands of the few to the many, democratizing the technology, and building trust in its use and future through transparency.” And for many developers, that may be enough.

Meta certainly knows – really open source or not – that being open will help its product. After all, as Nick Clegg, chairman of global affairs at Meta and former British deputy prime minister, said on BBC Radio 4’s Today, open source would make Llama-2 “safer and betterBy using the “wisdom of the crowds, you make these systems safer and better and, most importantly, you get them out of the… sweaty hands of the big tech companies who are currently the only companies that have either the computing power or the vast reservoirs of data to build these models in the first place.”

Eric S Raymond, author of the founding open source book The cathedral and the bazaarcould have written this.

But the devil is in the details when it comes to open source. And there Meta, with its Llama 2 Community License Agreement, comes crashing down on its head.

As The register noted earlierTHE community agreement prohibits the use of Llama 2 to train other language models; and if the technology is used in an application or service with more than 700 million monthly users, a special license is required from Meta.

Stefano Maffulli, Executive Director of OSI, explained: “While I’m glad Meta is pushing the bar for available access to powerful AI systems, I’m concerned about the confusion of some who celebrate LLaMa 2 as open source: if it were, it would have no restrictions on commercial use (points 5 and 6 of the open-source definition). As it stands, the terms applied by Meta allow only certain commercial use. The key word is some.”

Maffulli then dived deeper. “Open source means that developers and users are able to decide for themselves how and where to use the technology without needing to engage with another party; they have sovereignty over the technology they use. When read superficially, Llama’s license says, ‘You can’t use it if you’re Amazon, Google, Microsoft, Bytedance, Alibaba or your startup gets that big.’ That might sound like a reasonable clause, but it also implicitly says, “You have to ask us for permission to create a tool that could solve world hunger” or something important like that.”

Stephen O’Grady, open source licensing expert and co-founder of RedMonk, explained it this way: “Imagine if Linux was open source unless you worked at Facebook.” Exactly. Maffulli concluded, “That’s why open source has never put restrictions on the domain of use: you can’t know in advance what may happen in the future, good or bad.”

The OSI isn’t the only open source savvy group dealing with the Llama 2 license. Karen Sadler, an attorney and executive director of the Software Freedom Conservancy, dug into the language of the license and found that “additional commercial terms in Section 2 of the license agreement, which limit the number of users, make it nonfree and nonopen source.”

For Sadler, “It looks like Meta is trying to push a license that has some pitfalls of an open source license but, in fact, has the opposite result. Additionally, the Acceptable Use Policy, which the license requires compliance with, lists prohibited behaviors that are very broadly written and could be enforced in a very subjective way – if you send a mass email, could that be considered spam? If there is reasonably critical material posted, would it be considered defamatory?”

Last, but not least, she “did not notice any public drafting or commenting process for this license, which is necessary for any serious effort to introduce a new license”.

We asked Meta for comment. ®

Leave a Reply