OpenAI Drops 'Politically Unbiased' AI Reference in Policy Update

OpenAI Revises Stance on AI Political Bias

OpenAI has recently made a subtle change to its published policies, removing language that previously advocated for “politically unbiased” artificial intelligence.

The initial version of the company’s “economic blueprint” for the AI sector within the United States stated that AI models “should aim to be politically unbiased by default.”

Policy Document Update

A revised draft, released on Monday, no longer includes this specific phrasing.

Responding to inquiries, an OpenAI spokesperson explained that the alteration was implemented to “streamline” the document.

The spokesperson further indicated that other OpenAI documentation, notably the OpenAI’s Model Spec, adequately “make[s] the point on objectivity.”

The Model Spec, published in May, provides insights into the operational behavior of OpenAI’s diverse AI systems.

The Political Landscape of AI Bias

This revision also underscores the increasingly complex political debates surrounding “biased AI.”

Numerous allies of President-elect Donald Trump, including Elon Musk and David Sacks – a figure involved in both crypto and AI – have alleged that AI chatbots are suppressing conservative perspectives.

David Sacks has specifically criticized OpenAI’s ChatGPT, claiming it is “programmed to be woke” and presents inaccurate information on politically charged topics.

Attributing Bias in AI

Elon Musk attributes the issue to both the datasets used for AI training and the prevailing ideologies within San Francisco Bay Area tech companies.

“A lot of the AIs that are being trained in the San Francisco Bay Area, they take on the philosophy of people around them,” Musk stated at an event supported by the Saudi Arabian government last October.

He characterized this as a “woke, nihilistic – in my opinion – philosophy” being integrated into these AI systems.

The Technical Challenges of AI Bias

However, bias within AI represents a significant and persistent technical challenge.

Even xAI, Musk’s own AI company, has encountered difficulties in developing a chatbot free from political leanings.

Research published in August by U.K.-based researchers suggested that ChatGPT exhibits a liberal bias when addressing subjects like immigration, climate change, and same-sex marriage.

OpenAI maintains that any observed biases in ChatGPT are unintended “bugs, not features.”