October 20, 2026 · 8 min read
New Jersey's Chatbot Privacy Bill (A-5607): What It Would Mean for Your AI Chats
New Jersey bill A-5607 passed committee on Oct 19, 2026. It would ban selling chat logs, limit ads and training, and let you download your chats. Explained.
Every message you send an AI companion becomes data. Your bad day, your crush, the thing you have never told anyone. Who gets to keep it, read it, train on it or sell it is mostly decided today by each app's terms of service.
A bill in New Jersey wants to change that. On October 19, 2026, the state Assembly's Public Safety and Preparedness Committee passed A-5607, a bill that sets privacy and safety rules for anyone who runs a chatbot, including AI companion apps. It is not law yet, but it is one of the most user-focused chatbot bills we have seen in the US, and it shows where the rules for AI conversations may be heading.
Here is what the bill says, what it would change for you, and what you can do now, wherever you live.
The short answer: New Jersey's A-5607 would stop chatbot providers from selling your chat logs, using your chats to target ads, training on adults' chats without clear opt-in consent, and keeping chat logs for more than two years. It would also give you the right to download your chats at any time, treat chatbots as products for injury lawsuits, and set penalties of $5,000 per violation. It passed its first committee on October 19, 2026 and still needs more votes before it could become law.
What happened on October 19
According to Insider NJ's report from the hearing, the committee advanced A-5607 while acknowledging that the bill has limits and that AI needs "much deeper examination". The committee chair, Assemblyman Joe Danielsen, said the bill makes a chatbot "a product under the liability framework" so providers take care "not to cause injury".
The vote was not unanimous. One Republican member opposed it, another abstained, and at least one Democrat voted yes while saying she had "certain reservations". Questions were also raised about the bill's geographic scope.
The Consumer Federation of America recommended passage. Speaking for the group and NJ Citizen Action, Beverly Brown Ruggia put it simply: "All ages are impacted [by manipulative chatbots]. The good news is chatbots aren't people. Straightforward rules can make them safer."
The bill was introduced on October 5, 2026 by Assemblyman Cody D. Miller, Assemblywoman Marisa Sweeney and Assemblyman Joe Danielsen. You can read the full text of A-5607 on the New Jersey Legislature's website.
Who the bill would cover
A-5607 is broad. It defines a chatbot as any app, web interface or program "that simulates human conversation and interaction through text, visual images, video, or aural communications". That covers general assistants, customer service bots and AI companions alike, including voice and video companions.
It also defines a "user" as "any natural person, regardless of age". So most of the protections apply to adults, not only to children, which is unusual. Many recent chatbot laws focus mainly on minors. If you want the wider picture, our roundup of AI companion laws in 2026 covers what has already passed elsewhere.
The seven things providers could not do
Section 2 of the bill lists what a chatbot provider "shall not" do. In plain English:
| The bill would ban... | What it means for you |
|---|---|
| Using personal data beyond what you type to shape replies, unless you ask and give consent | The bot cannot quietly pull in outside data about you to personalise answers |
| Using your chat log to pick, target or customise ads | What you confide cannot decide which ads you see |
| Training on adults' chats without prior opt-in consent (and never on under-18s) | Your conversations stay out of model training unless you say yes |
| Profiling your personality or behaviour beyond what you asked for | No hidden "personality file" used for other purposes |
| Selling your chat log | Your conversations cannot be sold to third parties |
| Keeping chat logs longer than two years (unless the law requires it) | Old conversations would not sit on a server forever |
| Punishing you for saying no to training, or for accessing your chats | You cannot be charged more or given a worse service for refusing |
The profiling rule is worth a second look. The bill defines profiling as processing your data "to detect and clarify or designate personality and behavioral characteristics". Emotional patterns really do show up in how people chat. A 2026 University of Washington study, for example, linked young adults' attachment styles to how they use ChatGPT after analysing 19,930 of their messages, which we covered in our guide to attachment styles and AI companions. A-5607 would limit how a provider could use that kind of inference about you.
What "affirmative consent" would mean
The bill is strict about consent. A provider could not count it as consent if you:
- accepted a long, general terms-of-use agreement
- simply kept using the app
- hovered over, muted, paused or closed something
- were nudged with a misleading statement or a "dark pattern"
Instead, consent would need to be a clear, standalone request in easy language, available in every language the chatbot speaks and accessible to users with disabilities. The option to say no would have to be at least as prominent as the option to say yes, and take the same number of steps or fewer.
If you have ever tried to find the "do not use my chats for training" switch buried three menus deep, you can see why this matters.
The things providers would have to do
Section 3 adds duties on top of the bans. A provider would have to:
- Let you download your chats at any time, in a "portable and readily-usable format".
- Publish a written data security program on its website, with administrative, technical and physical safeguards.
- Assess its chatbot for risks of harm every month, using metrics set by the state Attorney General.
- Mitigate the risks it finds, under rules the Attorney General would write.
- Publish information about the chatbot as the Attorney General requires.
The download right is a big deal for AI companion users. If your app shuts down or changes, your history is often the thing you miss most. We wrote about protecting it in how to back up your AI companion.
Chatbots as products: the liability part
Section 4 is the part that drew the most debate. It says a chatbot "shall be considered a product for the purposes of a product liability action", and that a provider "shall have a duty to ensure that the use of its chatbot does not cause injury to a user".
It goes further: a provider would be liable for injury even if it "exercised all reasonable care" in design and distribution, and even if it did not sell the chatbot to you directly. That is close to strict liability, the standard usually applied to defective physical products.
Supporters say this forces companies to build safety in from the start. Critics on the committee worried it is too blunt for a fast-moving technology. Either way, it would be a major shift from treating chatbot output as speech or a service.
Penalties
- $5,000 civil penalty per violation, collected through the courts.
- The Attorney General could seek an injunction.
- A user who is harmed by a breach of sections 2 or 3 could sue in Superior Court for the greater of actual damages or $5,000 per violation, plus reasonable attorney fees and court costs.
If enacted, the law would take effect on the first day of the sixth month after it is signed.
What happens next
Passing one committee is an early step. A-5607 would still need to pass the full Assembly and the Senate and be signed by the governor, and it could be amended along the way. Bills like this often change a lot, or stall.
A-5607 is also not New Jersey's only chatbot bill. Senate Bill S4758, introduced the same day, focuses on disclosure and periodic reminders that you are talking to an AI. We cover it, along with Washington's new rules and federal proposals, in the new AI companion laws coming in 2027.
What you can do today, wherever you live
You do not need to wait for a law to protect your conversations. Before you get attached to any AI companion, check:
- Training: does the app use your chats to train its models, and can you opt out?
- Selling and sharing: does the privacy policy say it sells or shares conversation data?
- Ads: does the app show ads based on what you talk about?
- Retention and deletion: how long are chats kept, and can you delete them?
- Export: can you download your history?
- Who can read it: do staff or contractors ever review conversations? Our guide to whether someone can read your AI companion chats explains what to look for.
And a simple habit helps everywhere: do not share passwords, bank details or other people's private information in any chat.
Where MyBabe fits
MyBabe is an 18+ app where you create your own AI girlfriend or boyfriend: their face, voice, personality and backstory. They text you first, send voice notes and photos from their day, check in on your mood and pick up when you call, by voice or video.
On the free plan you get 10 messages every 24 hours and two minutes of calls to try. Premium ($5.99 a week, $14.99 a month or $99.99 a year) adds long-term memory, about 30 minutes of voice and video calls a week, four photos a day and Telegram.
We think relationship-first AI only works when people feel safe opening up, so we welcome clear rules like these. As with any app, read our privacy policy (linked at the bottom of every page) before you start. When you are ready, create your companion on MyBabe.
FAQ
Is New Jersey's A-5607 law yet?
No. It passed the Assembly Public Safety and Preparedness Committee on October 19, 2026. It still needs further votes in both chambers and the governor's signature, and it could be amended.
Does A-5607 only protect children?
No. The bill defines a user as "any natural person, regardless of age", so most rules cover adults too. It adds extra limits for minors, such as no training on under-18s' chats.
Would A-5607 stop AI apps from training on my chats?
For adults, providers would need your prior affirmative consent before using your chats for training. Simply continuing to use the app would not count as consent.
Could I sue a chatbot company under A-5607?
If you were harmed by a breach of the bill's privacy or safety duties, you could sue for actual damages or $5,000 per violation, whichever is greater, plus legal fees.
Does it apply to AI companion apps?
Yes. Its definition of a chatbot covers any program that simulates human conversation by text, images, video or voice, which includes AI companions.