As Larry Ellison bets the farm, Oracle says it loves AI-written code, just not in OpenJDK
Oracle has told contributors to OpenJDK, the open-source Java project it stewards, that they must not submit code or other content generated by AI tools. The interim policy, published in a post tagged "legal" while a fuller policy is drafted, cites concerns over reviewer workload, safety, security and intellectual property, given that OpenJDK underpins mission-critical systems used by organisations worldwide. The move sits awkwardly alongside Oracle's own enthusiastic promotion of AI-written code within its products, highlighting an apparent contradiction in the company's stance on AI-generated software.
Under the new rules, contributions to OpenJDK Git repositories, GitHub pull requests, emails, wiki pages and Java Bug System issues must not include material generated in whole or in part by large language models or similar deep-learning systems, though developers may still use such tools privately to comprehend, debug, review code and conduct research. Oracle warned that "plausible-looking but incorrect code" could jeopardise safety and security. This contrasts with comments from CTO Larry Ellison, who has said Oracle's own AI models now write much of the company's code, and co-CEO Mike Sicilia, who touted AI coding tools as enabling smaller teams to work faster. The apparent inconsistency comes as Oracle plans to invest $70 billion in datacentre expansion over the coming year, up from $55.7 billion in fiscal 2026, a move that prompted S&P to downgrade its credit rating to BBB-, just above junk status.
In short
- Oracle bans AI-generated contributions to the OpenJDK project
- Policy contrasts with Oracle's heavy internal use of AI coding tools
- Comes as Oracle plans $70bn datacentre spend and faces a credit downgrade
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Oracle's argument is that OpenJDK is foundational infrastructure supporting mission-critical systems worldwide, and therefore warrants stricter standards than internal or commercial products. AI-generated code, however sophisticated, can harbour subtle bugs or security vulnerabilities that are difficult for human reviewers to catch at scale. Concerns about intellectual property provenance, reviewer workload, and the need for robust verification in a community-driven project are legitimate, and using AI tools privately for learning differs meaningfully from submitting AI-generated code into the public codebase.
The case against
Oracle's position appears inconsistent when weighed against the company's own significant investment in AI-written code and its executives' public statements that AI coding accelerates development. If AI-generated code is deemed sufficiently reliable for Oracle's commercial products and customers, the blanket ban on OpenJDK suggests the policy reflects different risk tolerances or governance concerns rather than technical objections to code quality. Furthermore, the prohibition may be overly cautious given rapid improvements in AI code quality and prevents the open-source community from adopting tools increasingly standard across the technology industry.
Explained simply
OpenJDK is the open-source project behind Java, a programming language used in many organisations’ important computer systems. Oracle stewards the project, while outside developers can contribute code and related material for others to review.
The debate is about whether contributors should be allowed to submit material produced by AI tools. AI can generate code that looks convincing but contains errors, so questions include how contributions are checked and who is responsible for them. Oracle’s own leaders have promoted AI tools for writing code, making the company’s approach to its public project part of a wider discussion about using AI in software development.