Oracle Bans AI Code in OpenJDK — but Not in Its Own Products

Oracle, which acts as the steward of the open-source OpenJDK project, has introduced a ban on accepting code generated by artificial intelligence. The reasons cited are risks to security and reliability, along with intellectual-property concerns.

The wording of the policy leaves developers some room to maneuver: using large language models privately — to debug or check your own code — is still permitted. What's forbidden is something else: submitting AI-generated fragments into the project's repositories, pull requests and other channels. In other words, a neural network is allowed as an adviser behind the scenes, but not as a co-author whose name ends up in the commit history.

Double Standards

What makes this ban most curious is the backdrop of what Oracle says about itself. Company co-founder Larry Ellison recently declared that Oracle's code is now written by AI models. He was echoed by co-CEO Mike Sicilia, who credited neural-network tools with the ability to ship products faster and with smaller teams of engineers.

The result is a stark contradiction: what's presented as a breakthrough and a way to cut costs for Oracle's internal development turns out to be an unacceptable risk for an open-source project under its own management. It's as if the steward of Java is telling the community, “Do as we say, not as we do.”

An Expensive Bet on AI

The financial side adds further context. This year Oracle is investing 70 billion dollars in expanding data centers — clearly betting on the wave of demand tied to artificial intelligence. The company's appetites haven't gone unnoticed: the rating agency S&P downgraded Oracle's credit rating to BBB-, just one notch above “junk” status, citing the uncertain returns on these investments.

So Oracle is simultaneously placing large sums on a future in which code is increasingly created by machines, while walling off one of the key open-source projects of the Java ecosystem from that very future. For the OpenJDK community, the decision makes sense in its own way — the questions of provenance and legal cleanliness of generated code still have no clear-cut answers. But at the corporate level, the gap between words and deeds looks especially glaring.

Source: Hacker News