Despite my own skepticism of AI, I use it a lot. ChatGPT, Grok, Muse, Google's Gemini, and Microsoft's Copilot have all been turned to for help or information from time to time. I don't like that some topics are off-limits, but a lot of those are things I wasn't going to talk to AI about in the first place. Still, I might like to chat up ChatGPT about the latest about the firearm industry from time to time.
And Everytown doesn't like that.
It seems that in addition to an assault weapon ban, trying to restrict AI chatbots from talking about guns is super high on Everytown's post-midterm agenda.
Everytown for Gun Safety, the national gun control group founded and funded by billionaire Michael Bloomberg, celebrated antigun Democrat midterm victories during a Nov. 5 virtual townhall briefing and outlined its next moves. Everytown intends to defend bans on commonly owned Modern Sporting Rifles (MSRs) at the U.S. Supreme Court, mobilize activists to shape public opinion and press artificial intelligence (AI) companies to restrict firearm-related content.
The disclosure about AI is particularly notable. Weeks after Everytown’s affiliated publication criticized firearm-related YouTube content, the Bloomberg-backed gun control organization confirmed it is in direct discussions with major AI companies about their policies.
...The briefing’s most consequential disclosure came from Everytown Labs Senior Director Sri Madhavan and Policy Advocacy Director Krystal LoPilato. They discussed the organization’s report, Artificially Assisted Gun Violence: Chatbot Risks and Preventative Steps for Responsible AI Companies, and LoPilato confirmed that several leading AI companies have met with Everytown to review its recommendations. Talks are continuing.
Those recommendations include limits on certain firearm-related content and question responses, human review of conversations suggesting violence or self-harm and stronger procedures for referring serious threats to law enforcement.
Preventing violent crimes and responding to credible threats are legitimate safety objectives. The concern is how these companies distinguish criminal assistance from accurate information about legal activity.
Bingo.
See, as it stands, AI chatbots are really like a well-read kid in the special ed class. They know a lot and have access to more information, but they often can't really think in a meaningful way. I call the moments when this happens "Turing fails," named after the Turing Test that would determine true artificial intelligence. Basically, can a chatbot convince someone they're talking to a person instead of a machine?
This matters because these chatbots cannot necessarily ascertain differences between big talk and terrible deeds, for one thing. Hell, I'm not sure that any of the ones I use could tell the difference between a discussion of stopping ability for self-defense purposes and the same discussion being had with a would-be mass killer.
Understand that I'm not taking the threat to push an assault weapon ban lightly. I just don't think they'll get one before Viramontes is decided, and I don't know that they'll be able to do anything about it after the decision drops.
But bullying AI companies into compliance bothers me on a visceral level, because we already know that AI has biases. Those biases are the result of the programmers and what information they determine to be credible enough for the chatbots to use. The New York Times, for example, is considered more authoritative on guns than, say, Bearing Arms, even though we report the truth and they report nonsense as established fact all the time.
Now Everytown is hoping to bully — and let's understand, this isn't some good-natured request being granted but likely something that comes with the threat of Everytown going to war against companies that have enough controversy to deal with — these companies into turning their chatbots into anti-gun tattletales just because they don't like that you might ask questions they don't like while picking a home defense firearm.
If I didn't despise them before...
