MorseSign inStart freeOpen Morse

New Jersey's AI chatbot bill, A5607: what it would do with your chat logs

The Morse teamPublished 11 min read

As of November 19, 2026, New Jersey's AI chatbot bill, A5607, hasn't become law. It cleared the Assembly Public Safety and Preparedness Committee on October 19, 2026 and now waits in the Assembly Appropriations Committee. If it passed as written, chatbot providers could not use your chat logs to target ads, could not sell them, would need your opt-in before training on them, could not train on a minor's chats at all, and could keep a chat log for no more than two years. It would also make a chatbot a "product," so a provider could be liable for an injury even when it took "all reasonable care" (bill text).

We read the bill as introduced, its status page, its committee statement, and its fiscal estimate on the New Jersey Legislature's site on November 19, 2026, along with the January 2026 People-First Chatbot model bill and EPIC's testimony, and the privacy pages of ChatGPT, Gemini, Claude, and Meta AI. Newsweek reported the bill on November 17, 2026. The model's authors deserve credit for writing rules about the chat log itself, which most chatbot laws never touch. This isn't legal advice.

What is New Jersey's A5607?

An eight-page Assembly bill whose synopsis reads "Establishes content and data privacy requirements for chatbot providers" (New Jersey Legislature, bill text). Its primary sponsors are Assemblyman Cody D. Miller (District 4), Assemblywoman Marisa Sweeney (District 25), and Assemblyman Joe Danielsen (District 17), with Assemblyman Anthony Angelozzi as co-sponsor. It was introduced on October 5, 2026.

The Legislature's status page lists two actions: introduced and referred to the Assembly Public Safety and Preparedness Committee on October 5, then "Reported and Referred to Assembly Appropriations Committee" on October 19. The committee vote was 5 yes, 1 no, and 1 abstention, with the chair, Danielsen, voting yes. EPIC wrote on October 20 that Appropriations would take the bill up "on Thursday" (EPIC), but as of November 19 the status page shows no action there and no Senate version.

From here it needs Appropriations, a vote of the full Assembly, the Senate, and the governor's signature. If it became law, section 8 says it would "take effect on the first day of the sixth month following the date of enactment," and the attorney general could take "any anticipatory action" before then (A5607). The Office of Legislative Services' fiscal estimate, dated October 27, calls the cost to courts and the Department of Law and Public Safety "indeterminate" and notes the executive branch didn't answer its request for a fiscal note.

Which chatbots would the New Jersey bill cover?

Nearly all of them. Section 1 defines a chatbot as "a software application, web interface, or computer program that simulates human conversation and interaction through text, visual images, video, or aural communications." A provider is anyone "who creates, distributes, or otherwise makes available a chatbot," and a protected user is "any natural person, regardless of age" (section 1).

There are no carve-outs. Most of the state chatbot laws in our state-by-state survey are written for companion bots, and several exclude customer service bots outright. A5607 would reach general assistants, support bots, and companion apps alike.

What would A5607 do with your chat logs?

The bill defines a chat log as "any input data, outputs generated by a chatbot, or record of the input data or outputs from user interactions with a chatbot," so your messages, the replies, and any record of either all count. Here is each chat-log rule, the section that sets it, and what four of the biggest chatbots' own pages say on that point, all read November 19, 2026.

The bill's ruleSectionChatGPTGeminiClaudeMeta AI
No using a chat log to decide whether to show an ad, which one, or how2(b)Free and Go plans show ads chosen partly on "the context and intent of your current conversation" (OpenAI)"Your Gemini Apps chats are not being used to show you ads" (Google)"Claude will remain ad-free" (Anthropic)"Depending on your region, your interactions with AI products can be used to personalize content and ads" (Meta)
No training on an adult's chats without "prior affirmative consent"2(c)(3)"Data sharing is enabled by default" for training on Free, Plus, and Pro (OpenAI)"Chat content including text & attachments" is "on by default if you are over 18" (Google)Used for training if "You choose to allow us," or if a chat is flagged for safety review (Anthropic)"Meta uses your interactions with AIs to improve AI at Meta," with no switch on that page (Meta)
No selling a chat log2(e)"We don't 'sell' Personal Data" (OpenAI)"We do not sell your personal information to anyone" (Google)Anthropic "does not 'sell' your personal data as that term is defined by applicable laws" (Anthropic)"We don't sell any of your information to anyone, and we never will" (Meta)
Keep a chat log no longer than two years, unless law requires it2(f)Chats "remain saved to your account until you delete them" (OpenAI)Auto-delete after 18 months by default, changeable to 3 or 36 months or off (Google); reviewed chats kept "up to three years" (Google)Up to 5 years, de-identified, with training allowed; flagged chats' inputs and outputs up to 2 years (Anthropic)Deleting messages, threads, or your account "may not delete our copy" (Meta)
Let you download your chat logs "at any time"3(b)Export requested from settings, with a download link (OpenAI)"You can also export your information," through Google Takeout (Google)Export from Settings, then Privacy, on the web or desktop app only (Anthropic)Access "in a portable format," depending on where you live (Meta)

Two more rules sit beside these. Section 2(c)(4) and 2(d) bar profiling "beyond what is necessary to fulfill an express request," and section 2(a) bars using personal data from outside the chat to shape replies unless you asked for it and gave affirmative consent. Section 2(g) bars charging more or serving you worse for refusing training or for downloading your logs (section 2).

To count as "affirmative consent" under the bill, a request must come in a "standalone disclosure," refusing must take "the same number of steps or fewer" as agreeing, and consent can't be "inferred from the inaction of the user" or from accepting "general or broad terms of use." As we read it, a training switch that starts on wouldn't qualify. Our survey of which AI companies train on your data lists the switches app by app. Gemini's row changed this week: Google's help page said its training settings were "off by default" when we read it on November 18, 2026, and on November 19 it says chat content training "is on by default if you are over 18" (Google).

Would the bill let you delete your chats?

Not by itself. A5607 has no deletion right of its own, and neither does the January 2026 model, which gives only a right to access your logs. EPIC's summary of the federal People-First Chatbot Act, introduced July 9, 2026 by Representatives Valerie Foushee and Greg Casar, says that version "gives users the right to access and delete retained chat logs" (EPIC).

New Jersey takes a different route. Section 1 of A5607 says chatbot providers "are controllers for the purpose of" the New Jersey Data Privacy Act, P.L.2023, c.266, which gives consumers the right to "delete personal data concerning the consumer" (New Jersey Legislature). That act applies only to controllers that handle the data of at least 100,000 consumers a year, or 25,000 if they profit from selling data, and A5607 doesn't say whether its sentence removes that threshold. The privacy act also says it creates no private right of action.

Section 3(b) may pull the other way. It says a provider "shall" "retain in a portable and readily-usable format and make available for a user to access and download at any time any of the user's chat logs" (bill text). The model bill's version covers only logs "that a chatbot provider has retained." As we read it, New Jersey's wording could be taken as a duty to keep chat logs, though it sets no minimum period, while section 2(f) caps the period at two years. The bill doesn't say how the two fit together, or how either fits with a deletion request.

What does A5607 say about children's chats?

It protects two age groups and requires no age check. Under section 2(c), a provider may not process the chat log or personal data of a user it "knows or should know" is under 13 without a parent's or guardian's affirmative consent. It may not use a chat log or personal data for training at all if it knows or should know, "based on the knowledge fairly implied on the basis of objective circumstances," that the user is under 18 (section 2).

The model bill left the parental-consent age in brackets for each state to fill in, "[13 or 18]." New Jersey chose 13.

Who would enforce it, and could you sue?

Both the state and harmed users. Section 5 sets a civil penalty of "$5,000 for each violation," collected under New Jersey's Penalty Enforcement Law of 1999 in Superior Court or municipal court, and lets the attorney general seek an injunction. A user "who has been harmed by a violation" of the data or website rules in sections 2 or 3 can sue in their county's Superior Court for "the greater of actual damages or $5,000 for each violation," plus attorney fees and court costs (section 5).

Section 4 is the product-liability part. "A chatbot shall be considered a product for the purposes of a product liability action," a provider has "a duty to ensure that the use of its chatbot does not cause injury to a user," and it is liable for an injury even if it "exercised all reasonable care" or never dealt with the user directly. Section 7 adds that where A5607 conflicts with another New Jersey law, "the law that affords the greatest protection from harm to users shall control" (bill text). The attorney general would write rules for the monthly risk assessments and the website disclosures in section 3.

How is A5607 different from EPIC's People-First model bill?

It copies most of the model's data rules word for word and leaves out several of its other sections. The People-First Chatbot Bill, written by the Consumer Federation of America, EPIC, and Fairplay and dated January 2026 (model text), is the model EPIC says A5607 is based on. EPIC Counsel Kara Williams, one of the model's authors, testified for the bill on October 19 (EPIC testimony).

ProvisionModel bill, January 2026A5607 as introduced
Longest a chat log may be kept"longer than 10 years" barred"longer than two years" barred
Police and government accessNo compelled access "except as pursuant to a wiretap warrant"No such section
Notice that the bot isn't humanBefore any output, "every hour thereafter," and when askedNo such section
Claims of professional adviceNo implying output comes from a licensed doctor, lawyer, accountant, or financial planner, or that chats are "confidential"No such section
Your copy of the logsAccess to logs "that a chatbot provider has retained"Provider "shall" retain them for download
Parental consent age"[13 or 18]," left to the state13
Who can bring a caseAttorney general, "a district attorney, or a municipality"Attorney general (injunction) and a $5,000 civil penalty
User lawsuitsA violation "constitutes an injury in fact"; punitive damages for reckless violations"A user who has been harmed"; no punitive damages named
Tie to state privacy lawNoneProviders are "controllers" under the New Jersey Data Privacy Act

Newsweek's report says the bill would "prevent chatbot providers from presenting chatbot outputs as professional advice" (Newsweek). That rule is in the model's section 3, but the version of A5607 on the Legislature's site has no such section, and the committee statement doesn't list one either (committee statement). The bill could still be amended in Appropriations.

Do other states have chatbot laws like this?

Not with these chat-log rules. Of the 24 state chatbot laws we read on November 13, 2026, none bars a company from training on an adult's chats or caps how long it keeps them, and only Utah's mental health chatbot law bars selling what you type. The full table is in AI chatbot laws by state. EPIC also backed a second New Jersey bill on chatbot liability, A5090, which the Assembly Science, Innovation, and Technology Committee discussed on October 19, the day A5607 cleared committee (EPIC).

How Morse keeps your chat history in your browser instead

Every rule in A5607 limits what a company may do with chat logs it already holds on its servers: what ads they feed, whether they train a model, whether they're sold, and how long they sit there.

Morse doesn't train AI on your chats, sell your data, or show ads. Its servers don't keep messages or log message content, and your account holds sign-in details, plan, balance, and a usage ledger with no content. Your chat history is stored in your browser, encrypted, not on Morse's servers.

Each message still reaches a model, and the mode you pick decides who handles it:

  • Private, for chats you'd rather no model host kept at all. It's the default for a new chat. Open-weight models like DeepSeek, Kimi, and GLM, plus Grok, answer under Morse's zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
  • E2EE, for the chats you'd least want in anyone's log. Your browser checks the enclave's hardware attestation, then encrypts the message so only that enclave can read it, and Morse's servers relay text they can't read. It rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models with no web search or image generation, and comes with Pro at $19 a month.
  • Anonymous, for GPT, Claude, or Gemini without your account attached. Our proxy removes your account, email, and IP address and sends the message under Morse's own account, so the provider sees it without knowing it's you. The provider may keep it for a limited time to check for abuse, under its own policy, so leave out names and details that identify you.

The model bill's warrant section never made it into A5607. Because we don't store chats, we usually have very little to give authorities, as our page on whether AI chats can be subpoenaed explains. Clearing your browser, losing the device, or deleting your account deletes your chats, and Morse can't recover them, so keep a copy of anything you need. Morse is for adults only: every user confirms they're 18 or older. More on how this works is on our privacy page.

Send a first message from the Morse home page without an account, or start free with no card.

Common questions

Would A5607 cover ChatGPT, Gemini, and other general assistants?

As written, yes. Its definition of a chatbot is any program that 'simulates human conversation and interaction,' and unlike several state chatbot laws, it has no exception for customer service bots.

Is there a federal version of the People-First Chatbot bill?

EPIC says Representatives Valerie Foushee and Greg Casar introduced a People-First Chatbot Act in Congress on July 9, 2026, built on the same model. EPIC's summary says the federal bill would give users the right to access and delete retained chat logs, which New Jersey's bill doesn't do on its own.

Would the bill make chatbots check my age?

No. A5607 has no age verification duty. Its rules for minors apply when a provider 'knows or should know' a user is under 13 or under 18, so they turn on what the company knows rather than on a check.

When would it take effect if it passes?

On the first day of the sixth month after the governor signs it. The attorney general could start writing rules before then, since section 8 allows 'any anticipatory action' needed to implement it.

Ask what you actually want to ask.

80+ models, characters, and images, with every chat encrypted and kept on your device. Free to start.

Start a private chat

Read next