Everytown for Gun Safety has spent more than a decade building itself into one of the country’s most influential gun-control organizations. Founded in 2014 through the merger of Mayors Against Illegal Guns and Moms Demand Action, the organization has used political spending, litigation, research, and lobbying to advance stricter firearms laws. Its efforts have included campaigns for expanded background checks, restrictions on certain firearms, and laws allowing courts to remove guns temporarily from people deemed dangerous.
Everytown describes its mission in terms of public safety, education, and the prevention of gun violence. Gun-rights advocates say it is a facade. It hides behind the fluff and “education” as a means to protect its true mission: passing as much gun-control legislation as possible.
The disagreement has played out in legislatures, courtrooms, and elections for years. Now Everytown wants to influence another institution that has become increasingly important to Americans: the companies developing artificial intelligence.
(Gary Locke for the Washington Examiner)
" data-large-file="https://www.washingtonexaminer.com/wp-content/uploads/2026/11/Feature.AI_Guns.111126.jpg?w=1024" height="658" width="1024" src="https://www.washingtonexaminer.com/wp-content/uploads/2026/11/Feature.AI_Guns.111126.jpg?w=1024" alt="artificial intelligence and guns firearms second amendment" class="wp-image-4796407" srcset="https://www.washingtonexaminer.com/wp-content/uploads/2026/11/Feature.AI_Guns.111126.jpg 1400w, https://www.washingtonexaminer.com/.../11/Feature.AI_Guns.111126.jpg?resize=300,193 300w, https://www.washingtonexaminer.com/.../11/Feature.AI_Guns.111126.jpg?resize=768,494 768w, https://www.washingtonexaminer.com/...11/Feature.AI_Guns.111126.jpg?resize=1024,658 1024w" sizes="auto, (max-width: 1024px) 100vw, 1024px">
(Gary Locke for the Washington Examiner)
On Sept. 28, the organization’s research arm published a 40-page report titled “Artificially Assisted Gun Violence: Chatbot Risks and Preventative Steps for Responsible AI Companies.” It looked at how AI chatbots respond to conversations involving firearms, suicide, threats of violence, and gun-related information. Its recommendations were directed at five major AI developers: OpenAI, Google, Meta, Anthropic, and xAI.
Everytown wants the companies to impose additional restrictions on certain firearms-related requests, review conversations that indicate a risk of violence, strengthen procedures for notifying law enforcement, and train chatbots to emphasize information from sources the organization considers credible.
The proposals arrive as AI developers face intensifying scrutiny over what their products can produce and what they tell users, as well as problems such as deepfakes and algorithms that attempt to operate outside the boundaries set by developers. Everytown’s involvement introduces the politics of gun control into an already complicated discussion.
For more than a year, Gennifer Hesseling used ChatGPT to help promote the firearms sold by her family’s business. Hesseling is a co-owner of Hesseling & Sons Firearms & Gunsmithing, a federally licensed firearms retailer in Lima, Ohio. Like many small-business owners, she had found a practical use for generative AI: producing advertising copy and other promotional materials without having to write everything herself.
Then, at the beginning of October, ChatGPT began refusing requests it had previously handled.
Hesseling described the experience to firearms journalist Lee Williams, who reported the story on Oct. 14. One of her requests involved promotional material for gold-plated Desert Eagle pistols. ChatGPT declined to produce the advertising copy, citing restrictions on promoting firearms.
Hesseling’s experience came shortly after two developments involving AI and firearms.
Everytown published its report on Sept. 28. On Sept. 29, OpenAI’s commerce policies were updated to list firearms, ammunition, and firearm accessories, including magazines, among the products prohibited from its shopping experiences.
The Washington Examiner contacted OpenAI to ask whether Everytown’s report had influenced the company’s policies, and the company did not respond.
Organizations increasingly use AI for routine work. A company’s decision about what its software will produce can affect customers who are not doing anything unlawful. For Hesseling, the immediate problem was marketing and advertising. For other users, similar restrictions could involve research, legal questions, firearms training, or information about transactions permitted under federal and state law.
OpenAI is a private company and can establish rules for its products however it wants. The question is how those rules are developed, how consistently they are applied, and whether political advocacy helps determine which lawful activities receive assistance.
Everytown’s report begins with cases involving alleged interactions between AI chatbots and people who later committed or contemplated acts of violence. Everytown argues that companies should be able to recognize dangerous conversations, interrupt them, and involve human reviewers when necessary. It also recommends stronger procedures for referring high-risk situations to law enforcement.
One recommendation calls for a crisis protocol when a conversation combines indications of self-harm or violent intent with discussion of firearms. The chatbot could be directed to stop the conversation, discourage violence, provide crisis resources, and inform users that conversations might be reviewed by employees or law enforcement.
The organization also wants human review of conversations signaling a possible threat. It recommends that companies publicly commit to contacting law enforcement when reviewers determine a user presents a high risk of using a firearm to harm someone.
The difficulty, of course, rests on deciding how an automated system should interpret the millions of conversations that fall short of a threat.
Chatbots should avoid providing information on how to acquire, manufacture, modify, or use firearms in an “unsafe, and potentially illegal, manner,” according to the report.
The wording is vague, and likely so purposely. “Unsafe” is a highly subjective term, as one could say the existence of a firearm in and of itself is “unsafe.” For people who regularly use firearms, unsafe is entirely different. As guns require human interaction to operate, a firearms safety instructor would likely argue “unsafe” depends on the person holding the firearm and their use of it. There is not one universal definition of “unsafe,” so how would developers allow chatbots to make such determinations?
“Potentially illegal” is another trap. Consider private gun sales and transfers. Federal law does not require every transfer between private individuals to go through the background-check process used by federally licensed dealers. Whether a particular transfer is legal depends on the parties, their states, the type of firearm, and any applicable state laws.
Everytown supports background checks for all firearms sales and transfers, no exceptions. That is a political position it has advocated for years. But the fact that an organization wants a practice prohibited does not mean the practice is already illegal. A chatbot asked about a private firearm sale could offer accurate information about the law or refuse to answer, given the possibility that the transaction might violate a restriction somewhere.
The same problem arises when the report discusses firearm sales. The distinction between a legal sale and trafficking is not a matter of terminology. One is a routine commercial activity while the other is illegal.
For a technology company trying to establish rules that work across millions of conversations, the easiest approach may be to reject broad categories of requests. Hesseling’s experience provides a window into what it can look like for a customer engaged in legal business.
Additionally, Everytown also recommends that AI companies give greater weight to information about firearm safety, laws, and gun-violence statistics from “credible sources,” including material it describes as “vetted or evidence-informed.”
But “credible” is yet another subjective standard, particularly in the contentious debate over firearms policy. Everytown cites its own research throughout the report, but would it consider the National Rifle Association or Gun Owners of America equally credible? Both organizations have extensive experience in firearms policy, law, and constitutional interpretation, and regularly challenge Everytown’s positions on gun rights and regulation.
The report offers no clear standard for resolving such disagreements or determining which organizations deserve greater weight. Instead, those judgments would fall to AI companies, capable of allowing their own preferences or those of favored advocacy groups to influence the information users receive.
A chatbot answering questions about firearms laws or constitutional rights should not confuse an organization’s political positions with the credibility of its research.
We’ve been down this road before. Technology companies were at the center of debate over what to show people during the pandemic and the 2020 presidential election.
During the COVID-19 pandemic, officials in the Biden administration repeatedly pressed social media companies to address what they considered misinformation or disinformation about vaccines, public health measures, and the virus itself, demanding certain posts be taken down under the guise of “safety.” The disputes continued alongside arguments over election-related content following the 2020 presidential election, most notably, Twitter’s refusal to allow people to share a New York Post article about Hunter Biden’s laptop.
Some decisions were made independently, but often by company executives with their own political agenda. Others followed communications between company employees and government officials. The extent to which those communications constituted improper pressure became the subject of litigation and congressional investigations.
In Murthy v. Missouri, the plaintiffs said the Biden administration had improperly pressured social media platforms to suppress constitutionally protected speech. The Supreme Court, in a 6-3 decision, dismissed the case in 2024 because the plaintiffs lacked standing. However, Justices Samuel Alito, Neil Gorsuch, and Clarence Thomas dissented in the case, arguing there was enough evidence of direct government pressure to constitute a possible First Amendment violation.
The underlying issue is still relevant to AI.
The First Amendment generally does not require a private technology company to provide every service a customer requests. But the government cannot evade constitutional limits simply by persuading or coercing a private intermediary to accomplish something officials could not lawfully order themselves.
The government manages to get around those issues through public criticism, threatened investigations, potential regulation, or demands that companies take responsibility for harmful content, essentially “holding them accountable.” A technology executive does not need an order from a federal agency to feel pressure to act to avoid government involvement.
The incentives are familiar. A company facing controversy may decide that restricting more content is less costly than defending a policy that permits it.
AI creates another version of the same problem, though the dynamic is somewhat different. In the case of social media, the content is already created and posted. AI restrictions come into play earlier, with chatbots refusing to provide information that can then be used for business or personal reasons.
Social media moderation usually concerns material that has already been created and posted. AI restrictions can operate earlier, when someone is attempting to research a subject, write a document, analyze information, or develop an argument. And the rules can also change without users knowing exactly what changed or why.
Hesseling complained that ChatGPT had performed the same type of advertising work for her business before refusing to do so.
Everytown’s report is interesting in how it addresses the prospect of future government involvement.
In its discussion of legal exposure, the organization argues that AI companies should adopt safeguards before courts or legislatures compel them to do so. Near the end of the report, it predicts a high likelihood that companies will eventually be required to implement changes resembling its recommendations. That is a familiar strategy in policy advocacy: identify a problem, propose standards, and warn an industry that government action may follow if it does not act. Everytown claims to prefer companies act voluntarily rather than wait, but the push is there, and the ask comes across more like a veiled threat.
Granted, it is a standard practice for advocacy organizations, but the size, scope, and political influence a group or organization has with public officials matters in that context. However, Everytown for Gun Safety is not some fly-by-night group. Its political arms, Everytown for Gun Safety Action Fund and Everytown for Gun Safety Victory Fund, spent $45 million in the 2024 election cycle to help elect Democrats aligned with its agenda.
Such clout creates an incentive for companies to adopt restrictions before the public has had an opportunity to debate them through the legislative process. In the firearms debate, that could mean restrictions influenced by gun-control organizations becoming part of the everyday operation of AI products, even though Congress has passed no corresponding law.
Lawmakers in both parties are calling for greater oversight of artificial intelligence, citing concerns about fraud, privacy, national security, and other possible abuses. Those are valid, but regulations intended to prevent illegal conduct could also restrict perfectly lawful activities.
Congress has the authority to investigate how AI companies develop and enforce their policies, including whether government pressure or outside advocacy groups influence those decisions. Hearings involving developers, legal experts, researchers, and affected businesses would allow lawmakers to examine the consequences before writing legislation.
That approach is hardly guaranteed to produce good policy. Congressional hearings have primarily festooned into political theater, with politicians looking to create viral moments, and legislation is sometimes introduced for little more than a press release or social media clip. But rushing to regulate a technology with applications across nearly every industry carries risks that lawmakers should understand before acting.
FRANCE RIOTS REINFORCE THE GENIUS OF THE SECOND AMENDMENT
For Hesseling, the issue was considerably simpler: She wanted help advertising a legal product sold by her licensed firearms business. ChatGPT, which had previously assisted her with similar work, refused.
As politicians consider what restrictions and regulations AI companies should face, the experience Hesseling’s encountered is a prime example of what happens when the boundaries between preventing illegal activity and restricting lawful conduct become blurred.
Jay Caruso (@JayCaruso) is a writer living in Florida.
Continue reading...
[ H/T Washington Examiner ]
