The First Amendment and Artificial Intelligence

Should the First Amendment protect speech made by artificial intelligence?

My short answer: yes.

In March of 2016, Microsoft released an experimental chatbot on Twitter which they called MS Tay. Microsoft described MS Tay as an artificial intelligence program which would learn from and interact with other Twitter users and produce its own original content based on these interactions.

MS Tay was shut down less than 24 hours after the program had been implemented. Twitter users realized fairly early on that MS Tay’s program was easily manipulated, so of course the trolls took over, eventually causing MS Tay to spout antisemetic, misogynistic, and transphobic hate speech.

If MS Tay had been a real person, while Twitter could have removed this speech from their platform as they are a private company, the First Amendment undeniably would still protect their speech. However, as MS Tay is an artificial intelligence, whether the speech was protected is up for debate.

AIs are becoming more and more integrated into our everyday world. Businesses of all kinds are utilizing AI to do everything from bettering their company’s cybersecurity to evaluating their customer relationship management systems. The fast food giant McDonald’s has taken a recent and notable foray into AI when the burger chain acquired several artificial intelligence companies in what looks like an effort to eventually utilize artificial intelligence in automating their restaurants’ drive-thrus.

With today’s current technology, it’s hard to argue that the artificial intelligences created are advanced enough to have their speech be considered their own. The code which AIs are based on is written by human programmers, and therefore the expressions of these AIs can by extension be considered the expression of their programmers. But AIs are only becoming more and more sophisticated, and it’s likely that in our not-so-distant future there will be an AI so advanced that its statements and actions won’t be able to be considered attributable to its programmers.

So once we reach the point that an AI’s speech won’t be able to be attributed to any human, will that speech be protected by the First Amendment? Well, to take a textualist perspective of the Constitution, nowhere in the First Amendment does it say that it only protects human speech. To go further, the purpose of the First Amendment is not just to protect the speech rights of the individual, but it also protects the general public’s right to access information.

There are definitely issues with granting AI free speech rights, but they are pretty much the same issues we face when we give human beings free speech rights. And suppressing AI speech results in nearly the same harms that occur when we suppress human free speech.

Luckily, we’ve still got a ways to go before we can create an AI advanced enough that it can be considered able to think on its own. And even then, this conversation doesn’t have to be an “all or nothing” kind thing. Just like we suppress certain kinds of human speech to keep people safe (yelling “fire” in a crowded theater, threatening someone with violence, sharing top secret government information), these types of speech should also be suppressed when made by AIs.

Freedom of speech is a two-way street. It allows an individual to share their ideas as well as encounter the ideas of others, some of which may oppose their own (arguably the more important side of the street). This is true whether the speech be made by a human being or by an artificial intelligence, and this is why speech made by AIs should be protected under the First Amendment.