Google stopped its main AI chatbot, Gemini, from taking pictures of people. Users said the pictures were wrong or unfair to different races and ethnicities. They said some pictures it made were insensitive or showed harmful stereotypes about race.

---

• Google pauses new AI chatbot’s ability to generate images of people after backlash over racial bias.
• Decision made after users criticized insensitive historical depictions and stereotypes in images
• Pause shows the challenges of moderating AI tools that create novel content autonomously.

---

Google Pauses AI Image Generation of People After Backlash Over Race

One widely shared case showed Gemini generating an image of a racially diverse group of German soldiers in 1943 when the Nazi party was in power.

Other vital details on the situation

• Gemini refused requests to explicitly generate images of white people, citing concerns about harmful stereotypes. It did not appear to have similar restrictions on non-white ethnicities.
• The feature was flooded with examples of insensitive historical images after launching on the social network X last week.
• Google operates dozens of AI teams working semi-independently, complicating content moderation policies.
• Worker activism against unethical AI use has risen at Google and other tech firms.
• Congress has ramped up scrutiny of tech’s impact on children and vulnerable groups.

Google initially defended Gemini’s propensity to create racially diverse images reflecting its varied global user base.

But on Thursday, Google said Gemini still needed work on some race-related pictures.

This shows how hard it is for tech companies to check AI systems that can make new things independently. Other chatbots like OpenAI’s have had the same problems.

Experts say this happens because AI has training on vast amounts of data. Some of that data shows the biases and stereotypes in our society. Tech companies have to walk a fine line. They want AI to be creative. But they also have to stop it from making harmful stuff.

Google’s decision to halt image generation shows it is taking a cautious approach for now. Last year, Google said it would limit how its chatbots answer some questions about elections. But it did not give details.

If Google keeps the pause on Gemini making pictures of people for a long time, it could slow down Google’s new AI products. Google just launched a paid version of Gemini that can make advanced pictures.

It also integrated Gemini into its flagship Google search app on Android and iPhones.

Stopping this big feature of its new clever chatbot may upset users. They were excited to see what fun things it could create. But experts say Google made a smart move. They say it’s good that Google is taking time to improve Gemini’s ability to handle race carefully.

“It’s smart to pump the brakes and refine the system before relaunching this capability,” said Maribel Lopez, principal analyst at Lopez Research. “Google realizes that releasing an imperfect tool can damage its brand reputation.”

The path forward for Google and other chatbot creators essentially boils down to more caution, greater transparency, and exhaustive testing before launch.

According to Rashida Richardson, director of policy at AI Now Institute, Companies must embed ethics review and oversight into all stages of the AI development pipeline. The stakes are too high to leave these systems unchecked before unleashing them on billions of users.