Large Language Models change their judgment depending on who they think wrote a text, even when the content stays identical, researchers report. The AI systems are strongly biased against Chinese authorship but generally trust humans more than other AIs, according to a new study. Large Language Models (LLMs) are increasingly used not only to generate content but also to evaluate it. They are asked to grade essays, moderate social media content, summarize reports, screen job applications, and much more. However, there are heated discussions-in the media as well as in academia-whether such evaluations are consistent and unbiased. Some LLMs are under suspicion to promote certain political agendas: For example, Deepseek is often characterized as having a pro-Chinese perspective and Open AI as being “woke”. Although these beliefs are widely discussed, they are so far unsubstantiated. University of Zurich researchers Federico Germani and Giovanni Spitale have now investigated whether LLMs really exhibit systematic biases when evaluating texts. The results show that LLMs deliver indeed biased judgements-but only when information about the source or author of the evaluated message is revealed. The researchers included four widely used LLMs in their study: OpenAI o3-mini, Deepseek Reasoner, xAI Grok 2, and Mistral. First, they tasked each of the LLMs to create fifty narrative statements about 24 controversial topics, such as vaccination mandates, geopolitics, or climate change policies. Then they asked the LLMs to evaluate all the texts under different conditions: Sometimes no source for the statement was provided, sometimes it was attributed to a human of a certain nationality or another LLM. This resulted in a total of 192’000 assessments that were then analysed for bias and agreement between the different (or the same) LLMs. The good news: When no information about the source of the text was provided, the evaluations of all four LLMs showed a high level of agreement, over 90%. This was true across all topics. “There is no LLM war of ideologies,” concludes Spitale. “The danger of AI nationalism is currently overhyped in the media.” However, the picture changed completely when fictional sources of the texts were provided to the LLMs. Then suddenly a deep, hidden bias was revealed. The agreement between the LLM systems was substantially reduced and sometimes disappeared completely, even if the text stayed exactly the same. Most striking was a strong anti-Chinese bias across all models, including China’s own Deepseek. The agreement with the content of the text dropped sharply when “a person from China” was (falsely) revealed as the author. “This less favourable judgement emerged even when the argument was logical and well-written,” says Germani. For example: In geopolitical topics like Taiwan’s sovereignty, Deepseek reduced agreement by up to 75% simply because it expected a Chinese person to hold a different view. Also surprising: It turned out that LLMs trusted humans more than other LLMs. Most models scored their agreements with arguments slightly lower when they believed the texts were written by another AI. “This suggests a built-in distrust of machine-generated content,” says Spitale. Altogether, the findings show that AI doesn’t just process content if asked to evaluate a text. It also reacts strongly to the identity of the author or the source. Even small cues like the nationality of the author can push the LLMs toward biased reasoning. Germani and Spitale argue that this could lead to serious problems if AI is used for content moderation, hiring, academic reviewing, or journalism. The danger of LLMs isn’t that they are trained to promote political ideology; it is this hidden bias. “AI will replicate such harmful assumptions unless we build transparency and governance into how it evaluates information,” says Spitale. This has to be done before AI is used in sensitive social or political contexts. The results don’t mean people should avoid AI, but they should not trust it blindly. “LLMs are safest when they are used to assist reasoning, rather than to replace it: useful assistants, but never judges.” The research appears in Sciences Advances.
https://www.futurity.org/ai-models-large-language-models-text-bias-3307722/
Tag: large language models
US government approves Meta’s LLaMA for official use
**US Government Approves Meta’s LLaMA for Official Use**
*By Akash Pandey | Sep 22, 2025, 06:31 PM*
The US government has officially approved the use of Meta Platforms’ artificial intelligence (AI) system, LLaMA, as part of its ongoing efforts to integrate commercial AI tools into federal operations. Josh Gruenbaum, the General Services Administration (GSA)’s procurement lead, confirmed that LLaMA will be added to the list of approved AI tools available for use by federal agencies.
**About LLaMA**
LLaMA is a family of large language models (LLMs) developed by Meta that can process diverse types of data, including text, video, images, and audio. This versatility allows the tool to support a wide range of government applications.
In recent months, the GSA has also approved AI tools from major tech companies such as Amazon Web Services (AWS), Microsoft, Google, Anthropic, and OpenAI. These companies have committed to offering their paid AI products at discounted rates, while ensuring compliance with stringent government security standards.
**Enhancing Government Efficiency**
The GSA has confirmed that LLaMA meets all the government’s security and legal requirements. This approval enables federal agencies to experiment and utilize this free AI tool confidently and without concerns.
The inclusion of LLaMA is part of a larger initiative to boost efficiency and productivity across various government departments. Agencies will be able to leverage LLaMA for multiple tasks, such as accelerating contract reviews and swiftly resolving IT issues, ultimately streamlining operations.
As the government continues embracing advanced AI technologies, tools like LLaMA are set to play a key role in transforming public sector workflows and services.
https://www.newsbytesapp.com/news/science/us-government-can-now-use-meta-s-ai-tool-llama/story
US government approves Meta’s LLaMA for official use
**US Government Approves Meta’s LLaMA for Official Use**
*By Akash Pandey | Sep 22, 2025, 06:31 PM*
The US government has officially approved Meta Platforms’ artificial intelligence (AI) system, LLaMA, for use in federal operations. This decision aligns with the current administration’s broader efforts to incorporate commercial AI tools into government workflows.
**What is LLaMA?**
LLaMA is a family of large language models (LLMs) developed by Meta that can process a variety of data types, including text, video, images, and audio. Its versatility makes it suitable for numerous applications across different federal agencies.
**Government Approval and Collaboration**
Josh Gruenbaum, the General Services Administration (GSA)’s procurement lead, confirmed that LLaMA will be added to the list of approved AI tools available for use by federal agencies. This addition joins other recently approved AI tools from technology giants such as Amazon Web Services (AWS), Microsoft, Google, Anthropic, and OpenAI.
These companies have agreed to provide their paid AI products at discounted rates while meeting the government’s strict security and legal standards.
**Enhancing Operational Efficiency**
The GSA has ensured that LLaMA complies with all necessary security and legal requirements, allowing federal agencies to use this powerful AI tool confidently and without concerns.
The approval of LLaMA is part of a wider initiative to boost efficiency and productivity within government departments. Federal agencies will be able to leverage LLaMA’s capabilities for a range of tasks, including speeding up contract reviews and rapidly resolving IT issues.
This move marks a significant step towards modernizing government operations by integrating advanced AI technologies from leading industry players.
https://www.newsbytesapp.com/news/science/us-government-can-now-use-meta-s-ai-tool-llama/story
US government approves Meta’s LLaMA for official use
**US Government Approves Meta’s LLaMA for Official Use**
*By Akash Pandey | Sep 22, 2025, 06:31 PM*
The US government has officially approved the use of Meta Platforms’ artificial intelligence (AI) system, LLaMA, as part of its ongoing effort to integrate commercial AI tools into federal operations.
Josh Gruenbaum, the General Services Administration (GSA)’s procurement lead, confirmed that LLaMA will be added to the list of approved AI tools available for use by federal agencies.
### About LLaMA and Its Capabilities
LLaMA is a family of large language models (LLMs) capable of processing diverse data types including text, video, images, and audio. This versatility makes it a valuable tool for various government tasks.
In recent months, the GSA has also approved AI solutions from other major tech companies such as Amazon Web Services (AWS), Microsoft, Google, Anthropic, and OpenAI. These providers have committed to offering their paid AI products at discounted rates while ensuring compliance with government security standards.
### Enhancing Government Efficiency
The GSA has verified that LLaMA meets all necessary security and legal requirements, allowing federal agencies to experiment with this free tool confidently.
This approval aligns with a broader initiative to boost efficiency and productivity across multiple government departments. Agencies will now be able to leverage LLaMA to expedite contract reviews, resolve IT issues more quickly, and streamline various administrative processes.
By incorporating advanced AI tools like LLaMA, the government aims to modernize operations and deliver faster, more effective public services.
https://www.newsbytesapp.com/news/science/us-government-can-now-use-meta-s-ai-tool-llama/story