LLVM declares its relicensing effort complete
More than 99.5% of code ever contributed is now under Apache-2.0 with the LLVM exception, has left the tree, or is too small to copyright.
LLVM is calling its multi-year relicensing finished. More than 99.5% of all code ever contributed is now covered by Apache-2.0 with the LLVM exception, has left the code base, or is too small to be covered by copyright. "We think that's a good moment to call the effort complete," the project wrote.
A single license removes a lingering compliance burden for anyone who ships LLVM. A small slice of older code could still sit only under the project's historical terms, and product teams had to account for that slice in legal reviews and software bills of materials even when it added little practical risk. With project-owned code under one license, and exceptions marked, those reviews get simpler to state and easier to audit.
The work began in 2018, when LLVM started asking past contributors to relicense. New contributions have arrived under Apache-2.0 with the LLVM exception since January 19, 2019. Code from before that date was under the earlier UIUC/NCSA license, with MIT on some runtime libraries, so moving it required author permission.
More than 1,300 individual contributors and about 100 companies and organizations signed agreements covering their historic work. Coverage was measured by matching those agreements against the monorepo history and sorting each commit as covered, not relevant, or uncertain. Contributions of 10 lines or fewer were treated as too small for copyright, and code that is no longer in the tree was set aside.
What remains is a thin residue, not a second license regime. For companies embedding LLVM, the practical change is that the old dual-track accounting can stop.