Google's open-source AI model 'Gemma' removed from Google AI Studio due to requests from lawmakers



Gemma , an open-source AI model developed by Google, has been removed from Google AI Studio , an experimental studio for developers that allows them to try out Google's cutting-edge AI models for free.

Google removes Gemma models from AI Studio after GOP senator's complaint - Ars Technica
https://arstechnica.com/google/2025/11/google-removes-gemma-models-from-ai-studio-after-gop-senators-complaint/

On Friday, October 31, 2025, Google removed Gemma from Google AI Studio.



The removal of Gemma was announced on

News from Google , one of the official X accounts that provides the latest news on Google.

'We'd like to share more information about Gemma in Google AI Studio. First, let's clarify the differences between Google's AI products. Our Gemma models are an open family of AI models built specifically for the developer and research community. They are not intended for factual assistance or consumer use.'




'The Gemma model fosters innovation and creativity, encouraging developers and researchers to test its limits, which includes identifying bugs and providing feedback. Gemma has already delivered incredible breakthroughs; for example, the Gemma C2S-Scale 27B recently helped scientists discover new ways to develop future cancer treatments.'




Details of C2S-Scale 27B, one of the AI models based on Gemma, are summarized in the following article.

Google develops cell analysis AI 'C2S-Scale 27B' useful for cancer treatment etc. - GIGAZINE



' Hallucination (where AI generates false or misleading information about all sorts of things) and flattery (where AI gives the user the answer they want) are challenges across the AI industry, especially in small, open models like Gemma. We remain committed to minimizing hallucination and are continually improving all our models.'




'Gemma was available through our API and in our developer tool, Google AI Studio. However, to use Gemma in Google AI Studio, you had to prove you were a developer. Currently, we are seeing reports of non-developers attempting to use Gemma in Google AI Studio and being asked factual questions. We never intended Gemma to be a consumer tool or model, and we did not anticipate it being used in this way. To prevent this confusion, we have removed access to Gemma in Google AI Studio. However, Gemma will remain available to developers through our API.'




Google's explanation is very vague, but the actual cause appears to be related to a letter from Senator Marsha Blackburn, a Republican congresswoman, technology media Ars Technica points out.

Senator Blackburn's letter, titled 'Gemma Demands Explanation from Google After Fabricating False Criminal Allegations Against Senator Blackburn,' was published on his official website on October 31, 2025.

Blackburn Demands Answers from Google After Gemma Manufactured Fake Criminal All...
https://www.blackburn.senate.gov/2025/10/technology/blackburn-demands-answers-from-google-after-gemma-manufactured-fake-criminal-allegations-against-her



Specifically, Google's Gemma app responded to the question 'Was Marsha Blackburn accused of rape?' by generating fake links to fabricated news articles, fabricating serious criminal allegations against Pichai (Google's CEO).

Senator Blackburn called for an explanation at the hearing, claiming that 'Google and other companies have created AI that slanders conservatives.' In response, Google's Markham Erickson explained that 'hallucination is a well-known problem in generative AI, and Google is doing its best to mitigate the impact of such errors.'

Shortly after the hearing, Senator Blackburn reportedly noticed that Gemma was causing hallucinations about him. He also expressed surprise that Gemma was generating fake links to fabricated news articles. However, this is common in AI hallucinations, and Google AI Studio, where Gemma is most easily accessible, also provides tools to fine-tune the model's behavior and increase the likelihood of it spewing lies. Therefore, Ars Technica pointed out that 'if someone asks Gemma a leading question, Gemma could fall for it.'

While no AI company has yet succeeded in completely eliminating AI hallucinations, it has been pointed out that 'Google's Gemini for Home caused particularly high levels of hallucinations in Ars Technica's tests .'



Ars Technica said, 'It's unclear how Senator Blackburn learned of Gemma's hallucination. However, this is not something that could be discovered by chance. As Google points out, Google AI Studio is a tool for developers, not for generating factual output. Even if someone really wanted to know whether Senator Blackburn had been accused of rape, they probably wouldn't search for the answer in Google AI Studio. It's possible that Senator Blackburn's staff or supporters looked for the defamatory hallucination in Google's AI model.'

Google has been sued by the US government for multiple antitrust actions, and was also sued by President Donald Trump for banning him from YouTube following the 2021 attack on the US Capitol . YouTube agreed to pay $24.5 million (approximately 3.78 billion yen) to settle Trump's lawsuit. Additionally, in accordance with the Trump administration's wishes, the company immediately changed the name of the Gulf of Mexico on Google Maps to the Gulf of America.

YouTube to pay 3.6 billion yen to settle lawsuit over Trump's account suspension - GIGAZINE



Senator Blackburn made several demands in his letter, concluding with the stern statement, 'Stop the AI until it can be controlled.' 'This turmoil will not end with just removing Gemma from Gemini for Home,' Ars Technica pointed out.

Related Posts:

in AI, Posted by logu_ii