Google's AI Governance Plan Draws the Boundaries of What Counts as Harm

Google has published a new AI governance position paper, 'A Pragmatic Approach to AI Governance in America.' In it, Google rejects both overly restrictive rules and an unregulated 'Wild West.' The company's stated aim is to define what counts as harm in AI systems.

The paper acknowledges that powerful models may assist serious cyberattacks or dangerous biological research. Google argues for a governance approach that targets these high-severity risks without imposing broad restrictions on AI development. The paper's position is that not all potential harms should be regulated equally. It proposes a middle path between strict pre-approval regimes and no oversight at all.

The proposal positions Google in the middle of the US debate over AI regulation. By naming serious cyberattacks and dangerous biological research as areas of concern, the company signals which risks it considers urgent. The article's title suggests the plan's central move is boundary-setting: determining what qualifies as harm and, by implication, what falls outside that category. The paper's 'pragmatic' framing appears intended to present Google as a moderate actor in that debate. Whether such boundaries would be operationalized in any binding form is not detailed in the article.