BLUF
In other words, look at how consistently inconsistent AI already is in its biases, without the intervention of powerful government actors. Imagine just how much more biased it can get — and how difficult it would be for us to recognize it — if we hand the keys over to the government.
A Tale of Two Congressional Hearings (and several AI poems)
We showed up to warn about threats to free speech from AI. Half the room couldn’t care less.
Earlier today, I served as a witness at the House Judiciary Committee’s Special Subcommittee on the Weaponization of the Federal Government, which discussed (among other things) whether it’s a good idea for the government to regulate artificial intelligence and LLMs. For my part, I was determined to warn everyone not only about the threat AI poses to free speech, but also the threats regulatory capture and a government oligopoly on AI pose to the creation of knowledge itself.
I was joined on the panel by investigative journalist Lee Fang, reporter Katelynn Richardson, and former U.S. Ambassador to the Czech Republic Norman Eisen. Richardson testified about her reporting on government funding the development of tools to combat “misinformation” through a National Science Foundation grant program. As FIRE’s Director of Public Advocacy Aaron Terr noted, such technology could be misused in anti-speech ways.
“The government doesn’t violate the First Amendment simply by funding research, but it’s troubling when tax dollars are used to develop censorship technology,” said Terr. “If the government ultimately used this technology to regulate allegedly false online speech, or coerced digital platforms to use it for that purpose, that would violate the First Amendment. Given government officials’ persistent pressure on social media platforms to regulate misinformation, that’s not a far-fetched scenario.”
Lee Fang testified about his reporting on government involvement in social media moderation decisions, most recently how a New York Times reporter’s tweet was suppressed by Twitter (now X) following notification from a Homeland Security agency. Fang’s investigative journalism on the documents X released after Elon Musk’s purchase of the platform has highlighted the risk of “jawboning,” or the use of government platforms to effectuate censorship through informal pressure.
Unfortunately, I was pretty disappointed that it seemed like we were having (at least) two different hearings at once. Although there were several tangents, the discussion on the Republican side was mostly about the topic at hand. On the Democratic side, unfortunately, it was overwhelmingly about how Trump has promised to use the government to target his enemies if he wins a second term. It’s not a trivial concern, but the hearing was an opportunity to discuss the serious threats posed by the use of AI censorship tools in the hands of a president of either party, so I wish there had been more interest in the question at hand on the Democratic side of the committee.