Oracle's OpenJDK bans AI-generated code contributions as the interim policy draws fresh scrutiny

Oracle’s interim policy for OpenJDK bans AI-generated content — including source code, text and images — from every contribution channel of the open-source Java project, a rule now drawing renewed developer scrutiny months after it was published.

The interim policy bars any contribution ‘generated, in part or in full, by large language models, diffusion models, or similar deep-learning systems’ across Git repositories, GitHub pull requests, emails, wiki pages and issue-tracker entries. Editing 10 of 100 AI-generated lines does not make a contribution acceptable; the whole submission remains disallowed.

Oracle cites three reasons: the review burden AI-generated code places on maintainers, safety and security in the mission-critical systems that run on Java, and unresolved intellectual-property and copyright questions around model outputs. Developers may still use large language models privately for debugging or review; they simply cannot submit AI-generated material to the project.

The policy is explicitly interim while Oracle drafts a full version for the OpenJDK Governing Board. Its renewed attention stems from the contrast with Oracle’s public posture: co-founder Larry Ellison has said AI now writes much of the company’s own commercial code, even as OpenJDK bars it outright.

The gap is less contradictory than it looks. Oracle’s commercial code runs through its own reviewers and liability regime, while OpenJDK is a public commons where a contributor’s provenance is hard to verify and IP exposure is shared. Still, the policy leaves open how maintainers would detect AI-generated submissions in the first place; enforcement rests largely on contributor honesty.