Should AI Have Free Speech?

Free speech is a privilege that many of us hold dear. But as machines become increasingly capable of generating words, ideas and arguments, should this fundamental right be extended to them too?

This might seem like a strange question at first. Machines don’t really have views and opinions, so why should they need the right to speak them?

But here’s the crux of the issue: Today’s intelligent machines create messages and content that can shape public opinion, perhaps even influencing politics and democracy.

So to what extent should those with the power to influence what they say—the AI companies that build them, and the governments that legislate how they work—control and shape their output?

As AI becomes increasingly ingrained in everyday life, this is a question that society is being forced to tackle.

It’s a complex issue central to the ongoing debate over how humans and intelligent machines will co-exist. So here I’ll explore both sides of the argument to understand why it matters in politics, our personal day-to-day lives and business.

Why Does AI Free Speech Matter?

Machines don't have or need the same rights as we do. So when we talk about AI and free speech, we’re really addressing the rights of those who control the output of machines.

This includes everyone who uses AI, but those with the greatest amount of power are probably the tech companies that create it and the governments that impose regulations on it.

From e-commerce recommendation systems to customer service chatbots and automated news feeds, what these systems say carries real weight, from influencing our buying decisions to defining the news agenda we’re exposed to.

Analysis of the large-language models powering ChatGPT-style chatbots, for example, finds that they often exhibit a left-leaning political bias.

This was noticed by Elon Musk, who said he’d address this by designing his own chatbot, Grok, to work as a “truth-seeking” engine. However, this resulted in a system that investigators say favors right-wing narratives.

So perhaps objective truth-seeking isn’t always as simple as it might seem.

With decisions made in governments, public policy think tanks and boardrooms increasingly leaning on AI-derived research, the political, business and societal implications are far-reaching.

Should we limit the extent to which we allow those with influence to steer AI’s output—impinging perhaps on their free speech—in order to protect the “freedom” of AI to give us unfiltered, objective facts?

Whose Free Speech Are We Protecting?

When AI “speaks”, whose voice are we hearing?

By upholding the rights of AI creators to influence the output of their models however they see fit, we could be simply handing corporations and governments megaphones to amplify their own views.

Here’s another example. Since it was launched, the output of ChatGPT and many other LLM chatbots has been moderated in order to prevent them from giving out harmful information as well as to prevent breaches of copyright laws. It won’t give out advice on committing crimes, for example, or let users create non-consensual deepfaked pornography.

These are examples of creators exerting control over AI output that many people might think are entirely sensible.

But what about when a government with a poor human rights record creates its own LLM chatbot, and possibly bans it from discussing human rights abuses, or stifles legitimate, anti-government democratic debate?

Some have suggested that Grok’s shift to the right may stem from as simple a change as instructing it to “consider politically incorrect responses”, rather than simply gathering its understanding from mainstream media outlets.

But even this change could have huge implications in terms of the influence on responses that users receive. It could still be considered a choice of the creator to steer the AI output away from making a truly neutral decision, towards expressing fringe or populist views, or those that go against established scientific or societal consensus.

What Does This Mean For The Future Of Human Free Speech?

Of course, all these questions around the intersection of human and machine free speech aren’t really about granting rights to robots. It’s about understanding how machines, and those that build them, will influence our thinking and decision-making in the years to come.

Resolving this will become even more urgent as we move into the age of agentic AI. We’re no longer dealing with simple chatbots but autonomous, always-on agents, much more able to influence our behavior in the long term, rather than one decision at a time.

Imposing appropriate guardrails, while respecting the rights of those who build and use AI to express ideas as they see fit, will undoubtedly be a tricky balancing act.

But it’s a key element of the wider challenge facing society as it attempts to benefit from the possibilities created by AI, while minimizing its potential to cause harm.

For businesses, this means taking responsibility for their “AI voice” in terms of what it says, as well as how it aligns with core values they want to project to customers.

For legislators, it may mean tackling questions around when AI can present information as fact, and when it should be made clear that its views are influenced by those of its developers.

And for everyone else, it means recognizing that the information we consume, share and base our decisions on may increasingly be shaped by robots, following instructions that aren’t always out in the open.