Open source has always walked a fine line between community spirit and commercial survival. With the rise of generative AI, that balance is becoming harder to maintain. For many independent developers, permissive licences no longer feel like a safeguard of openness, but a threat to sustainability.
Herman Martinus, founder of Bear Blog, is among those moving away from permissive licensing. His concerns stem from how easily AI-assisted development enables competitors to rebrand and resell open projects, often at the expense of the original creator.
The Rise of “Free-Ride Competition”
Martinus, described what he calls “free-ride competition.” He told AIM, “Someone forking the code, rebranding and releasing a hosted version of the same software that competes with Bear Blog.”
Though these forks had not yet toppled his project, the practice felt “exploitative and not aligned with the reasons I made the code available in the first place.”
The tipping point came when he noticed that one of the clones had been created with AI tools.
Martinus observed that it was clearly AI-assisted at first glance. Although the underlying code was nearly identical, the text had been revised by a model.
“This was the final straw for me, where it felt like free-ride competition was now too easy and didn’t require much technical skill,” he said.
His experience reflects a broader concern within open source. What was once a hub for collaborative innovation in the GenAI era has quickly become a breeding ground for clones.
What the Broader Landscape Shows
Across the industry, there are signs that developers and companies are making similar moves.
The Linux Foundation has argued that AI models, unlike traditional software, combine code, training data, weights, and documentation. This creates licensing challenges that existing open-source frameworks are ill-equipped to handle.
This uncertainty has left many contributors questioning whether their work is sufficiently protected in an era where models can remix and repurpose content with little regard for attribution.
Some high-profile projects have already been abandoned with permissive licences. Elastic shifted from Apache 2.0 to a more restrictive licence in 2021, citing cloud providers that re-hosted its software as competing services. The company framed the change as a defence of sustainability, though it sparked debate about whether such measures still qualified as “open”.
Others, like Timescale and Confluent, have adopted “source-available” models that allow transparency without giving competitors a free pass to monetise their work.
Rohit Vyas, director of solutions engineering and customer success for South Asia at Confluent, told AIM, “From a licensing perspective, I would say that Confluent has always maintained a separation between open source and proprietary [license].”
He explained, “If you [talk about] the proprietary services of Confluent, then the licensing regime is built to ensure that we provide the right mix of the software, which is rightly traceable, at the right price point and geographically, globally available, consistent.”
Vyas added that open-source projects already have governance mechanisms that determine who can contribute to and manage the software and its main code branches.
“If somebody has to do something with an open source project under the hood [using] Gen AI, then it is left to them. The law catches up with them sooner or later,” he said.
Placed against this backdrop, Bear CMS’s decision appears less like an isolated shift and more like part of an industry-wide recalibration. The idea that “open” automatically means permissive licensing is losing ground to a more defensive, sustainability-minded approach.
Restrictive Licences and the Future of Openness
The debate around restrictive licences is not new, but the AI factor has sharpened its urgency. Martinus sees the logic in the approach. “If it’s referring to a cohesive system that is the foundation of a company (and your livelihood), then it absolutely protects sustainability,” he said.
Yet he acknowledged trade-offs. His original motivation for releasing Bear’s code under MIT was transparency rather than adoption.
“The main reason I wanted to make the source available was to make my statements about privacy and security auditable,” he explained. That openness, however, became an unintended invitation for competition. “If I knew when I was starting what I know now, I would have made it source-available from the beginning,” he said.
Martinus noted that community backlash has been minimal, acknowledging that a small percentage—about 5 percent—of Hacker News users are likely to be upset by any issue. He dismissed the criticism, expressing the view that the decision will ultimately benefit the Bear project and will not significantly impact the open-source community.
Towards a Post-Open Source Era?
The bigger question is whether open source itself needs to evolve. Martinus was sceptical about the need for AI-era licences.
“We already know that AI companies do not respect licenses,” he said, pointing to controversies over dataset usage by Anthropic and Meta. “If you don’t want your code ingested, and want cloning of your software difficult, don’t release the code publicly.”
In his view, open source was facing strain long before AI. “Since the internet gained more commercial interest, there has been a tug-of-war between open-source development and commercial exploitation,” he said. AI, then, is only accelerating existing tensions.
And what of the future? Martinus remains uncertain. “If OpenAI is to be believed, all code will be written by AI in the future, and we won’t even need open-source software to build with. I remain sceptical, but I can’t predict the ever-more-uncertain future.”
As developers confront the new reality of AI-enabled cloning, the era of default permissive licensing may be giving way to more defensive strategies. Whether this protects innovation or fragments the ecosystem remains unresolved. But what is clear is that the GenAI era is forcing developers to rethink what openness really means.
The post GenAI Is Killing Old Open Source Rules appeared first on Analytics India Magazine.
