Apache 2.0 for Gemma: Build, Modify, Sell Freely
Gemma models grant perpetual, royalty-free copyright and patent licenses to reproduce, modify, distribute, and commercialize under Apache 2.0, requiring attribution retention, change notices, and license inclusion—ideal for production AI apps.
Usage Rights Unlock Commercial AI Builds
Apache 2.0 provides each contributor's perpetual, worldwide, non-exclusive, royalty-free, irrevocable copyright license to reproduce, prepare derivative works, publicly display/perform, sublicense, and distribute Gemma models in source or object form. Patent licenses cover making, using, selling, or importing the work, but only for claims necessarily infringed by the contributor's additions—terminate if you sue over the work. This setup lets you integrate Gemma into SaaS products, fine-tune for custom agents, or bundle in apps without royalty payments, as long as you comply with redistribution rules.
Redistribution: Four Conditions to Follow
Distribute unmodified or modified Gemma copies in any medium by: (a) including the full Apache 2.0 license; (b) adding prominent notices to changed files; (c) retaining all original copyright, patent, trademark, and attribution notices in source forms (omit irrelevant ones); (d) carrying over any NOTICE file contents in your derivatives via NOTICE file, docs, or UI displays. Add your own copyright or stricter terms to modifications, but never alter the original work's license. Contributions you submit default to Apache 2.0 unless specified otherwise. Trademarks like 'Gemma' can't be used beyond describing origin or NOTICE reproduction. These steps ensure legal forks, like RAG pipelines or hosted inference services, while protecting upstream contributors.
No Warranties, Your Risks
Gemma comes 'AS IS' without warranties of title, non-infringement, merchantability, or fitness—test thoroughly for production. Contributors limit liability for all damages (direct, indirect, etc.), even if warned. Offer paid support or warranties on your derivatives, but indemnify contributors. For AI builders, this means validate model outputs, handle hallucinations, and monitor costs yourself; the license shields Google and contributors from your app's failures.