Google Halts Gemini's Image Generation of People Over History Errors

Google has decided to pause, effective immediately, its Gemini model's ability to generate images of people. The move comes after users on X (formerly Twitter) began sharing screenshots in recent days showing the image generator producing glaringly inaccurate historical scenes: World War II German soldiers depicted with various ethnicities, and America's founding fathers—white men from the 18th century, according to the historical record—rendered with a racial diversity that doesn't match the era being portrayed.

Jack Krawczyk, senior director of product management for Gemini at Google, publicly acknowledged the issue. "We're aware that Gemini is offering inaccuracies in some historical image generation depictions," he wrote on X, adding that the team is "working to improve issues like this immediately."

An overcorrection problem, not censorship

The root of the error isn't a random technical glitch but a deliberate design decision. Like many generative image models, Gemini was tuned to avoid a well-documented bias in this technology: trained mostly on internet data, these systems tend to default to associating professions, roles or historical eras with white people, often men, reproducing stereotypes rather than correcting them.

To compensate, Google introduced mechanisms that automatically diversify depictions of people when the user doesn't specify particular traits. The problem arises when that diversification is applied to contexts where historical accuracy demands the opposite: a Nazi army in 1943, a 15th-century papal conclave, or the signing of the 1776 Declaration of Independence don't allow for the same demographic flexibility as a generic illustration of "a doctor" or "a lawyer."

The result has been a model that, in its attempt not to discriminate, ends up distorting verifiable historical facts—a failure that has drawn criticism both from those who see it as evidence of ideological bias baked into the AI's design and from those simply pointing out a product quality flaw.

Why it matters beyond Gemini

This episode comes barely two weeks after Google rebranded Bard as Gemini, folding image generation in as one of its flagship features to compete with rivals like OpenAI's DALL-E or Midjourney. The pause therefore affects a capability that had just been heavily promoted, and it exposes the tensions facing any company deploying generative AI at scale: balancing representation and accuracy without letting one priority swallow the other.

The case also illustrates a structural problem across the industry. Adjustments to these systems—whether through hidden instructions to the model, safety filters, or retraining—rarely distinguish well between "generating without bias" and "generating with accuracy," because both goals require different rules depending on the context of each request. A system calibrated to correct racial bias in contemporary scenes can fail spectacularly in historical scenes, and vice versa.

What Google has said and what's still unresolved

Google has not offered a timeline for restoring Gemini's ability to generate images of people, beyond Krawczyk's commitment to work "immediately" on a fix. In the meantime, the model remains available for generating images of objects, landscapes, or scenes without human figures.

The company faces a dilemma that isn't new but is a recurring one across the industry: any adjustment that reduces the risk of unwarranted demographic bias risks introducing, if applied without nuance, another kind of error that's just as visible—and, in this case, easily verifiable by any user with a basic grasp of history.