Oracle Bars AI-Generated Code From OpenJDK While Its CEO Touts 'Star Wars' Coding Gains

Overview

Oracle’s flagship open-source Java project, OpenJDK, is telling contributors they may not submit AI-generated code — even as Oracle’s own executives publicly celebrate the productivity gains AI coding tools are delivering elsewhere in the company. The contradiction, first documented by The Register in August, resurfaced after CEO Mike Sicilia’s remarks at Deutsche Bank’s 2026 Technology Conference on September 1, delivered against the backdrop of a year in which Oracle’s workforce shrank by 21,000 people.

What We Know

OpenJDK’s “Interim Policy on Generative AI”, dated April 9, 2026, states: “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,” according to OpenJDK. The rule covers more than code — it extends to “source code, text, and images in OpenJDK Git repositories, GitHub pull requests, e-mail messages, wiki pages, and JBS issues,” according to OpenJDK.

The policy was approved by the OpenJDK Governing Board as an interim measure while Oracle, the project’s corporate sponsor, drafts a fuller policy, according to OpenJDK. It permits contributors to use generative AI tools privately “to help comprehend, debug, and review OpenJDK code and other content, and to do research related to OpenJDK Projects,” so long as they “do not contribute content generated by such tools,” according to OpenJDK. There is no partial-edit exception: per OpenJDK’s own FAQ, generating 100 lines of code with an AI tool and hand-editing ten of them still counts as an AI-generated contribution, according to OpenJDK. Enforcement will run through Oracle’s Skara code-review tooling, which the policy says will “shortly” gain a pull-request checkbox requiring contributors to affirm compliance, according to OpenJDK.

OpenJDK cites three reasons for the restriction: reviewer burden, since AI tools make it easy to generate “large quantities of plausible-looking code… which is nonetheless incorrect”; safety and security, because the JDK “sits at the foundation of mission-critical systems in businesses, governments, and other organizations around the world” and “plausible-looking but incorrect code would put these critical properties at risk”; and intellectual property, since the Oracle Contributor Agreement requires contributors to own the rights to their submissions, which most generative AI tools’ training data cannot guarantee, according to OpenJDK.

That caution sits apart from how Oracle talks about AI-written code elsewhere. At Oracle AI World 2025, co-founder and CTO Larry Ellison said, “The code that Oracle is writing, Oracle isn’t writing,” adding, “Our AI models are writing,” according to The Register. Earlier in 2026, Sicilia said AI coding tools “would be a threat if we weren’t adopting them, but we are and very rapidly,” and that they were “enabling smaller engineering teams to deliver more complete solutions” for customers, according to The Register.

At the Deutsche Bank conference on September 1, Sicilia said he would once have dismissed today’s AI-assisted engineering productivity as “like Star Wars, futuristic stuff,” but called it “super real,” according to The Register. He argued the software industry would be “disrupted in a positive way” by AI and pushed back on what he called “SaaSpocalypse fears” that coding agents would strip value from business-application vendors, according to The Register.

Sicilia’s comments came as Oracle’s headcount contracted sharply. The company’s workforce fell from about 162,000 employees in May 2025 to about 141,000 in May 2026 — a drop of 21,000, with US headcount down 9,000 and international headcount down 12,000 — according to The Register, which cited Oracle’s annual report. Oracle has stated that “the adoption and deployment of AI technologies across our operations have resulted, and may continue to result, in reductions to our workforce,” and added that “we may initiate new restructuring plans in the future,” according to The Register.

Oracle’s AI push has had a rockier reception with customers elsewhere. A revamped support portal, which relaunched in late 2025 promising AI-powered interactions, streamlined navigation and better search according to Greg Parikh, Oracle’s vice president for information development and operations, instead drew complaints that the AI was “not working that well,” with support note numbers disappearing, automatic service requests no longer possible, and users describing wasted time on broken links and limited patch search, according to The Register.

Separately, Oracle told investors last month it would spend $70 billion on datacenter build-out in the coming year, up from $55.7 billion in fiscal 2026, after which S&P downgraded Oracle’s credit rating to BBB- — one notch above junk status — citing an uncertain path to profiting from the investment, according to The Register.

What We Don’t Know

Oracle has not publicly explained why it considers AI-generated code acceptable, even desirable, in its own commercial engineering while barring it from OpenJDK contributions; The Register said it asked Oracle about the discrepancy and had not received an answer as of its report, according to The Register. OpenJDK’s full, non-interim generative-AI policy has also not been published; the current rules remain explicitly temporary “until that policy is in place,” according to OpenJDK.

Analysis

Oracle is not alone in drawing this line. The Rust project adopted a formal LLM policy in August that permits AI-assisted work but bars AI-authored submissions, part of a broader pattern of open-source maintainers weighing reviewer burden against AI coding tools’ productivity claims. What distinguishes Oracle’s case is the gap between how it treats its own commercial codebase and how it treats the shared, foundational Java runtime it stewards. Ellison’s claim that Oracle’s product code is now substantially AI-written sits uneasily next to OpenJDK’s own reasoning that AI-generated code is too risky for software that “sits at the foundation of mission-critical systems in businesses, governments, and other organizations around the world.” Sicilia’s pitch — that AI coding productivity will let Oracle “disrupt” itself without succumbing to the SaaSpocalypse — is being made at the same moment the company’s headcount is shrinking and Oracle itself cites AI adoption as a factor in the reductions. Whichever story is closer to the truth inside Oracle’s own product teams, OpenJDK’s Governing Board has, for now, decided it isn’t willing to bet the world’s most widely deployed Java implementation on it.