Google AI Principles

This document sets out Google's principles for how it will build and use AI responsibly. Google commits to only releasing AI where the benefits clearly outweigh the risks, to actively working to prevent bias and harmful outcomes throughout a product's entire life, and to keeping humans involved in overseeing AI systems. It also commits to protecting privacy and security and to respecting intellectual property as part of its AI practices.

This document establishes Google's self-imposed governance commitments for the development and deployment of artificial intelligence systems across the full model lifecycle. Google commits to deploying AI only where likely overall benefits substantially outweigh foreseeable risks, and to employing rigorous design, testing, monitoring, and safeguards to mitigate unfair bias and unintended or harmful outcomes. The document further commits Google to implementing appropriate human oversight, due diligence, and feedback mechanisms aligned with both user goals and internationally recognized human rights principles, and to maintaining continuously adaptive safety measures from pre-launch through post-launch stages. Additional commitments cover promotion of privacy and security in AI practices, respect for intellectual property rights, and sharing of safety learnings with the broader AI ecosystem.

For an individual user, this document means Google has committed to actively testing its AI systems for bias and harmful outcomes before and after launch, to aligning those systems with human oversight and user goals, and to promoting privacy and security throughout its AI practices. Google's safety measures are expressly designed to adapt continuously rather than remain fixed, so the specific safeguards in place may evolve over time. Users can reference these stated commitments as Google's declared standard for how its AI products are governed.