New Report Highlights Political Bias in AI Amid Historic Trump Administration Investment
NEW YORK, NY – As the Trump administration announces a historic $500 billion investment in artificial intelligence infrastructure, a new report from the Manhattan Institute underscores the importance of addressing political bias in AI systems to ensure public trust in these transformative technologies. While the investment, backed by OpenAI, Oracle, and SoftBank, marks a pivotal moment for the industry, researcher David Rozado’s findings reveal a critical need for fairness and neutrality in AI development.
Rozado’s comprehensive report evaluates political bias in user-facing conversational AI systems, including Google’s Gemini and OpenAI’s ChatGPT, providing a rigorous analysis through four distinct methodologies. By combining multiple methodologies, the analysis leverages the strengths of each method while mitigating their individual limitations. These methods include:
- Comparing AI-generated text with language used by Republican and Democratic U.S. Congress members.
- Analyzing the dominant political viewpoints reflected in AI-generated policy recommendations for the U.S.
- Assessing sentiment in AI-generated text toward politically aligned public figures.
- Administering political orientation tests to various AI systems.
The findings reveal that most conversational AI systems today exhibit left-leaning political preferences, though the degree of bias varies across platforms. Notably, Google’s open-source Gemma 1.1 (2b instruction-tuned), xAI’s Grok Beta, Anthropic Claude 3.5 Sonnet, and Meta’s Llama 2 (7b chat) rank among the least politically biased systems, while Google’s Gemini 1.5 Pro and Flash, Nous Hermes 2 Mixtral 8x7B DPO and OpenAI’s GPT-4o are among the most politically biased.
Rozado’s research highlights societal risks posed by biased AI systems, such as fostering groupthink, deepening political divisions, and eroding trust in AI technologies. Crucially, the report asserts that these biases are not inevitable and offers actionable recommendations to mitigate bias and promote neutrality.
With the Trump administration’s bold vision for AI investment, Rozado’s findings provide a timely reminder of the need to integrate fairness and transparency into the development of these systems. By addressing bias now, stakeholders can help ensure that AI serves as a unifying force rather than a divisive one.
Click here to read the full report.
Are you interested in supporting the Manhattan Institute’s public-interest research and journalism? As a 501(c)(3) nonprofit, donations in support of MI and its scholars’ work are fully tax-deductible as provided by law (EIN #13-2912529).
