Everytown says AI companies have met with it over firearm chatbot safeguards
The group is targeting five model developers, hoping restrictions will carry into other applications. It has not identified which companies joined the discussions.
Loading page…
The group is targeting five model developers, hoping restrictions will carry into other applications. It has not identified which companies joined the discussions.
Listen to this story
Everytown’s proposals would shape both chatbot answers and escalation procedures: they call for refusing certain firearm instructions, responding to signs of self-harm or violence, and adding human review, reporting channels, and audits. The group says it has held ongoing discussions with AI developers, but no policy changes were announced and participants were not named. OpenAI, Google, Meta, Anthropic, and xAI were targets of the September white paper, not confirmed meeting participants, leaving the scope of any developer action unresolved.
The September 28 white paper seeks refusals for instructions on bypassing background checks, making unserialized firearms, and 3D-printing firearms.
Everytown also recommends crisis responses, human review of high-risk conversations, referrals, user-report analysis, and internal audits.
NSSF argues broad refusals could impede access to accurate information about lawful firearm conduct; this is its objection, not a confirmed outcome.
Everytown for Gun Safety says several leading AI companies have met with it about firearm-related chatbot safeguards, bringing its recommendations directly to developers. At a November 5 virtual town hall, the group described continuing discussions but announced no agreed policy changes, according to an account published by NSSF, the firearm-industry trade association.
NSSF published its latest account on November 18. Larry Keane, its senior vice president for government and public affairs and general counsel, also described the meetings in a November 14 Outdoor Wire article. Those accounts come from the same author, not independent confirmations of the discussions.
The discussions concern Everytown's September 28 white paper, Artificially Assisted Gun Violence: Chatbot Risks and Preventative Steps for Responsible AI Companies. It calls for changes both to what chatbots tell users and to how companies handle conversations suggesting danger. Its proposals extend beyond refusing answers: they include crisis responses, human review, reporting channels and internal audits.
The paper draws on company policies and court filings available as of August 18, 2026, and warns that policies or litigation may have changed since then. Its inventory of existing safeguards is therefore a dated baseline for the advocacy, not a current assessment of every company's practices.
At the town hall, Everytown Policy Advocacy Director Krystal LoPilato described the conversations as “ongoing, productive,” NSSF reported. She did not identify the participating companies. The five companies targeted by the white paper—OpenAI, Google, Meta, Anthropic and xAI—should not be read as a confirmed list of meeting participants.
Everytown Labs lead Sriraman Madhavan explained why the group chose those developers: they operate popular consumer chatbots and also build the underlying models used by other applications. His stated ambition was to influence products beyond the companies' own chat interfaces, through changes to the models those products rely on.
Our recommendations can really drive changes that carry through to every product built on them
Sriraman Madhavan, Everytown Labs lead, as quoted by NSSF
NSSF's objection is that some proposed refusals could sweep in information about lawful conduct. It argues that acquiring a firearm without a federal background check or making one for personal use is not inherently illegal. Its account notes that eligibility, circumstances, state law and other restrictions affect what is permitted.
Keane warns that broad restrictions could leave retailers, manufacturers and firearm owners unable to obtain accurate explanations of their legal obligations. That is the trade association's concern about how the proposals might be implemented, not an announced consequence of the meetings. Madhavan's remarks describe Everytown's intended reach; they do not establish that developers have accepted it.
Everytown's proposed role for chatbots is not limited to withholding information. The white paper also urges companies to provide safe-handling and secure-storage guidance, and information about removing firearm access when someone is in distress. Its requested changes cover both the answers users receive and the response when a conversation signals danger.
Loading discussion...
Join the conversation
Explain where you would draw the line.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.