Everytown's push to limit AI gun chatbots could silence lawful questions too
Everytown for Gun Safety is pressing OpenAI, Google, Meta, Anthropic, and xAI to restrict firearm-related answers from their AI chatbots, and the group's own briefing made clear it wants those limits built into the underlying models, not just one app's chat window.
The short version
At a November 5 virtual briefing, Everytown officials said their recommendations are aimed at the five companies because they control both popular consumer chatbots and the foundation models that power other products built on top of them. In plain terms: a restriction baked into the model could follow you into tools used by firearm retailers, manufacturers, and ordinary owners researching their own state's laws, not just into a single chatbot's refusal to answer a question.
The National Shooting Sports Foundation (NSSF), the firearm industry's trade association, reports that Everytown's white paper does flag real safety concerns, like preventing chatbots from helping plan violence or encouraging suicide. But it also asks AI systems to refuse questions about background-check requirements, unserialized firearms, and 3D-printed guns, topics that are not automatically illegal and that lawful owners legitimately ask about.
Key points
- Five companies targeted: Everytown's September white paper names OpenAI, Google, Meta, Anthropic, and xAI, chosen because their models underpin products beyond their own chatbots.
- Model-level reach: Everytown Labs' Sriraman Madhavan said the goal is changes that "carry through to every product built on" these companies' models, not just one app.
- Legal gray zones swept in: Federal law doesn't require a background check for every private transfer, and not every privately made firearm needs a serial number. NSSF argues a blanket AI refusal can't tell a law-abiding question from a criminal request.
- No commitments confirmed: Everytown's policy director said talks with several AI companies are "ongoing" and "productive" but declined to name which companies or describe any changes agreed to.
- Litigation cited, not proven: Everytown pointed to pending lawsuits alleging chatbots aided attackers or suicidal users. NSSF notes allegations in active litigation aren't established facts.
- A narrower path exists: NSSF points to its own Project ChildSafe program as an example of AI potentially connecting users to safe-storage and handling resources without treating lawful ownership information as off-limits.
Bottom line
This is still in the negotiation stage, not a finished policy, but it's worth watching. If model-level restrictions land the way Everytown wants, the practical effect could be that a retailer's customer-service bot or your own search for "do I need a background check for a private sale in my state" gets stonewalled along with genuinely dangerous queries. Keep an eye on which companies actually adopt these recommendations, and don't assume a chatbot's refusal to answer means the underlying activity is illegal.
Source: nssf.org