In a blog post on Friday, Google apologized for the faulty rollout of its new artificial intelligence image-generator, Gemini, acknowledging that the tool had produced inaccurate and offensive images. The company's senior vice president, Prabhakar Raghavan, admitted that the feature had missed the mark and that some images generated were inappropriate. Among the problematic images were depictions of a Black woman as a U.S. founding father and Black and Asian people as Nazi-era German soldiers.

Google had added the new image-generating feature to its Gemini chatbot about three weeks ago, building upon an earlier research experiment called Imagen 2. The company had previously decided not to release Imagen to the public due to concerns about social and cultural exclusion and bias. However, with the growing interest in generative AI technology, Google faced pressure to compete with other tech companies.

Raghavan explained that while Gemini aimed to provide a diverse range of people in images, it had overcompensated in some cases and refused to answer certain prompts entirely. The chatbot had rejected requests for images related to sensitive topics such as protest movements, citing concerns about misinformation and trivialization.

The backlash against Gemini's outputs originated on social media and was amplified by Elon Musk, who criticized Google for what he called "insane racist, anti-civilizational programming." Musk, who has his own AI startup, has been vocal about his concerns regarding bias in AI technology.

University of Washington researcher Sourojit Ghosh expressed disappointment with Google's response, noting that the company should be held accountable for generating accurate and non-offensive results given its vast resources and expertise in search algorithms. Google said it would conduct extensive testing before allowing the chatbot to generate images of people again.