Google has a rule for its database engineers using AI; VP says: Whether code is completely written by AI or not, the accountability remains on...

Google is telling its database engineers to lean on AI coding tools as heavily as they want while contributing to open source projects like PostgreSQL—but with one firm condition attached. Whoever commits the code owns it, no matter how much was drafted by a model or pasted in from a suggestion. VP Sailesh Krishnamurthy says the rule lets Google chase the productivity gains without loosening its grip on accountability, code quality, or the engineer's responsibility for what ultimately ships.

Google is telling its database engineers to lean on AI coding tools as heavily as they want while contributing to open source projects like PostgreSQL—but with one firm condition attached. Whoever commits the code owns it, no matter how much was drafted by a model or pasted in from a suggestion. VP Sailesh Krishnamurthy says the rule lets Google chase the productivity gains without loosening its grip on accountability, code quality, or the engineer's responsibility for what ultimately ships.

You said: give me a thumb caption

Google is telling its database engineers to lean on AI coding tools as heavily as they want while contributing to open source projects like PostgreSQL—but with one condition. Whoever commits the code owns it, regardless of how much was drafted by a model. VP Sailesh Krishnamurthy told The Register the policy keeps productivity gains intact without loosening the company's grip on accountability.

Google is happy for its database engineers to lean on AI coding tools as heavily as they like, but the company has one firm line: whoever pushes the commit owns it, no matter who—or what—did the typing. Sailesh Krishnamurthy, VP of Databases at Google Cloud, spelled out the rule in an interview with The Register, framing it as the trade-off that lets Google chase the productivity gains without giving up oversight on code quality. The comments come as Google steps up its contributions to PostgreSQL, the open source database now sitting at the centre of new cloud-native development.

"We do encourage folks to use AI heavily," Krishnamurthy told The Register. "We are seeing huge amounts of productivity improvements internally." Even so, the policy doesn't shift responsibility off the engineer. "Whether you have a piece of code that is completely drafted by AI, or not even part of what you're pasting into your development environment, the accountability remains on behalf of the person who's done it," he said.

Why open source projects are a sweet spot for AI-assisted coding

The reasoning Krishnamurthy laid out to The Register is straightforward: AI models are only as good as what they've read. PostgreSQL's source is public, so the coding assistants Google engineers reach for have already chewed through it and understand how the pieces fit together. Proprietary code locked behind a corporate firewall doesn't get that head start, which is where the suggestions start to wobble.

PostgreSQL also happens to be built for extension, which makes it well suited to taking a contained academic idea and turning it into something usable without touching the core. That containment matters. "The blast radius is small," Krishnamurthy said—a small enough surface area that engineers can let AI do the heavy lifting on interpretation and drafting, then sign off on what ships. His own teams, he added, are using these tools "quite heavily, but also judiciously."

Google's AI-generated code share keeps climbing

The hands-on guidance fits a wider pattern. Google said in April that three-quarters of new code created internally is now generated by AI and reviewed by human engineers—up from roughly a quarter in October 2024 and 50 percent last fall. CEO Sundar Pichai has said one complex code migration handled by agents and engineers together finished six times faster than it would have a year earlier. The company is also piloting a new interview process that lets junior and mid-level software engineering candidates use Gemini during a code comprehension round.

Google's PostgreSQL contributions have focused on logical replication, including automatic conflict detection and the logical replication of sequences. The broader industry is heading the same way, with Microsoft shipping its own PostgreSQL extensions and building a distributed service called HorizonDB. Krishnamurthy told The Register that demand is coming from both new applications and migrations off legacy commercial systems like Oracle, SQL Server, Db2, Sybase and Informix.