Oracle, the corporate sponsor of OpenJDK, has barred contributors from submitting code or other material generated by AI to the open source Java project, even as its own executives say AI now writes much of Oracle’s code.
“Contributions in the OpenJDK Community must not include content generated, in part or in full, by large language models, diffusion models, or similar deep-learning systems,” the interim policy says. The ban covers source code, text and images in OpenJDK Git repositories, GitHub pull requests, e-mail messages, wiki pages and issues in the JDK Bug System (JBS). The OpenJDK Governing Board approved the interim rules in early April, and Oracle says it will propose a full policy to the board “in due course.”
Private use is still allowed. Contributors may use generative AI tools “to help comprehend, debug, and review OpenJDK code and other content, and to do research related to OpenJDK Projects,” as long as they don’t contribute content those tools generated.
Three stated risks
The policy gives three reasons. The first is reviewer burden: AI tools make it easy to produce large amounts of plausible-looking code and tests that are wrong or hard to maintain, which drains reviewers’ limited time. The second is safety and security, because the JDK sits under mission-critical systems in businesses and governments. “Safety and security are paramount. Plausible-looking but incorrect code would put these critical properties at risk,” the policy says. The third is intellectual property, since the Oracle Contributor Agreement requires contributors to own the rights to what they submit.
The FAQ is strict on edge cases. As InfoQ reported, editing ten of 100 AI-generated lines still isn’t allowed, because the contribution is still partly AI-generated. Editor features such as spell-checking, auto-completion and refactoring are fine if they aren’t based on large language models. Contributors will soon have to tick a checkbox in Skara, OpenJDK’s pull request review system, to confirm they comply, though the project concedes that reliably telling human content from AI content “is impossible.”
Another Oracle-backed project went the other way. GraalVM, an Oracle Labs project that the OpenJDK Governing Board doesn’t govern, published a policy in mid-April allowing AI coding assistants for contributions, with disclosure encouraged but not required. Both projects use the same contributor agreement, InfoQ noted.
AI inside Oracle
The Register pointed out the contrast with Oracle’s own practice. “The code that Oracle is writing, Oracle isn’t writing,” co-founder and CTO Larry Ellison told Oracle AI World 2025. “Our AI models are writing.” Co-CEO Mike Sicilia said earlier this year that “the use of AI coding tools inside Oracle is enabling smaller engineering teams to deliver more complete solutions to our customers more quickly.”
Oracle has said it will invest $70 billion in the coming year on its datacenter build-out, up from $55.7 billion in fiscal 2026, which ended in May. S&P responded by cutting Oracle’s credit rating to BBB-, citing an uncertain path to making money from the investments, The Register reported. The Register said it had asked Oracle why AI-generated code suits its products but not OpenJDK contributions.
WebProNews is an iEntry Publication