I used to prefer permissive licenses and now favor copyleft

Open-source developers are re-evaluating permissive licenses like MIT and BSD versus copyleft licenses such as the GPL, weighing corporate adoption against long-term user freedoms. Advocates of copyleft argue that permissive licenses effectively subsidize big tech, enable closed forks, and weaken the software commons, while supporters of permissive terms emphasize maximal reuse, simpler commercial models, and fewer legal risks. The exchange also touches on edge cases like SaaS, AI training, and service-provider licenses (e.g., SSPL), highlighting how licensing choices can shape power dynamics between individual developers, large companies, and end users.

Scope of the debate

  • Thread centers on whether shifting from permissive to copyleft licenses better serves users, developers, and society, with substantial disagreement on both practical and ethical grounds.
  • Many comments treat Vitalik’s shift as part of a broader re‑evaluation of permissive licensing in the age of cloud platforms and AI.

Reasons given for preferring permissive licenses

  • Some developers care mainly about immediate users and downstream developers, not broader social effects; they want minimum friction and obligations.
  • Permissive bases enable closed-source add‑ons and tooling businesses that wouldn’t exist (or be viable) under strong copyleft; this is framed as expanding the ecosystem and accessibility for non‑programmer users.
  • Companies often forbid GPL; permissive licensing is seen as the only way to get widespread corporate adoption, contributions, and sponsorship.
  • Some view their code as an unconditional “gift to the commons” and see copyleft conditions as hypocritical or authoritarian.

Reasons given for preferring copyleft / critiques of permissive

  • Copyleft is framed as caring for users’ long‑term freedom: preventing popular improvements from disappearing into proprietary forks or embrace‑extend‑extinguish scenarios.
  • Permissive licenses are repeatedly described as subsidizing big business and “dead branches” of proprietary code that don’t return value to the commons.
  • Several argue that permissive licenses benefit large firms far more than small ones, who can’t compete when their own work is productized against them.
  • Objection: “no one is harmed by permissive licenses” is challenged with examples where users of proprietary forks lose the ability to inspect, modify, and repair.

GPL, legal uncertainty, and alternatives

  • Strong disagreements over how GPL/AGPL apply to linking, plugins, and “infection”; some describe the text as vague and litigation‑prone, others say myths and corporate FUD exaggerate the risks.
  • Some report that GPL scares corporate users; others would rather forgo that adoption than enable proprietary restriction of users.
  • MPLv2 is praised as a pragmatic middle ground; LGPL and its static-linking obligations are seen as awkward for languages with heavy codegen/templates.
  • SSPL and similar “anti‑cloud” licenses are discussed as attempts to defend against hyperscalers, though their status as “free/open source” is contested.

Licensing and AI training

  • Multiple comments argue that copyleft (as written) is ill‑suited to AI and SaaS; calls appear for a GPLv4 or variants with no‑AI or stronger network-use clauses.
  • Others respond that copyright and fair‑use doctrines themselves may be failing against AI, making license clauses hard to enforce.

Broader context: diffusion and “theft”

  • Historical industrial espionage and copyright avoidance are cited to argue that technology diffusion (often via “theft”) has always driven growth, complicating strict pro‑property or anti‑copy narratives.