If AI Is Biased, What Can Developers Do Next?

Imagine everyone treated each other with perfectly fair judgment. People expressed thoughts equally, never swayed by background or identity. Doesn’t that world sound enticing?

The truth is, the promise of a completely unbiased environment feels like a distant reality. In work, school, communities, and general life, everyone always has some form of predisposition shaped by culture and experience. It’s a natural way of thinking that helps us all make the best decisions.

In the AI space, the story of bias is no different. It’s likely unchanging and something we as users will always have to deal with. But with its presence continually rising, millions raise the fundamental question: when will it ever not be biased? 

According to experts, removing bias from AI is a complex ask. Because AI is trained to compute responses based on existing data, it is nearly impossible to retrain what the system might think. Given large language models (LLMs) are also highly unpredictable, it is hard to determine what exactly the robot will generate.

“You can measure bias because it’s an attribute of the output of LLMs. But you’ll never neutralize it, actually you don’t want to. The idea that we’ll one day ‘eliminate’ bias from AI is a fantasy, because every model reflects the people, systems, and choices behind it,” says Nicolas Genest, CEO of CodeBoxx.

This type of problem complicates efforts to create universally accepted standards for fairness in AI. What one group considers fair, another may view as biased. For example, perhaps one AI resume reviewer designed to ignore gender might still disadvantage women if it draws from male-dominant data. In this case, it’s not clear who gets to decide what’s “fair,” leading to subjective results.

Regardless of the consequences, that is where transparency in AI development becomes essential. Rather than striving for complete balance, developers should focus on making their AI models more understandable. This involves documenting the data sources, design choices, and challenges that often influence AI systems.

Genest adds, “This is why the answer can only be to prompt your way out of bias if need be. The way it can be done, and absolutely should be, is making those value decisions visible. Expose the assumptions. Document the trade-offs. Stop pretending the system is objective even though you wished it was. That’s real accountability.”

Practical steps exist for developers to mitigate bias, even if it cannot be erased fully. These include diversifying trained data, stress-testing outputs against harmful stereotypes, and publishing model documentation. Thorough audits and external oversight can also serve as guardrails against unchecked bias.

Recent advancements demonstrate both progress and limitations. OpenAI’s ChatGPT-5, for instance, reported that the system showed 30% less measurable political bias than its previous models, which was determined based on internal evaluations. While results were encouraging, it still underscored the need for independent verification.

Importantly, humans must carry part of the weight. Full trust in AI is dangerous, especially as these systems make critical decisions. Asking who trained the system, what data it used, and whose interests it serves is crucial for keeping AI development responsible.

Now the question is, where does this leave the future of AI? If AI can never truly be unbiased, what’s next for those who depend on it?

Like Genest might put it, the ones who are looking for success aren’t expecting AI to change overnight. And in fact, they aren’t projecting it to be fair anytime soon. But at the least, developers can work to regulate the technology and detect suspicions whenever it may arise.

As AI technology continues to widen, the pursuit of fairness in the industry will unfortunately not come to fruition. But instead of chasing neutrality, we as a society must work to challenge the debate head on. 

If the future of technology is to serve the people, then there’s much work we must do now. By facing AI openly, we can decide how AI responds to us.

Reference: https://internetsearch.co/2025/10/19/if-ai-is-biased-what-can-developers-do-next/