Trump was supposed to sign an AI executive order on Thursday. He didn’t. He pulled it at the last minute and told reporters he “didn’t like certain aspects” and was worried it could slow American AI down. It’s the same order that had been postponed, by Axios’s count, several times already.
Outside of the federal government, there was progress. California’s state Senate unanimously passed the strongest chatbot protections for kids in the country, Missouri’s legislature sent the governor a bill that makes it illegal to pass off an AI chatbot as a therapist, and Oakland Unified School District released a draft policy for the wild west of AI in its classrooms.
Same Story, Different Day (a slightly different version of SSDD). The federal conversation stalls and the people closest to actual kids just... keep working. Here’s what they did.
The most important thing you can do is talk to your family and friends about AI and technology. Here is a question to get the conversation going this week.
When something new and powerful shows up in your life, like a chatbot that can talk to you about almost anything, who should decide what’s safe and what isn’t? You? Your parents? Your teachers? The company that built it? The government? Why?
On May 19, the California Senate passed SB 1119 unanimously, 39-0. The bill, authored by Sen. Steve Padilla (D-San Diego), now heads to the Assembly. It builds on SB 243, which Padilla wrote last year and Gov. Newsom signed.
Here’s what SB 1119 would require of companion chatbot operators by July 1, 2027:
Annual child-safety risk assessments, documented.
Independent audits filed with the attorney general.
Crisis-response protocols.
Stricter default safety settings for minors, time limits, parental controls.
An end to targeted advertising to minors.
An end to selling or sharing minors’ data.
A private right of action so parents can sue.
Maria Raine, mother of 16-year-old Adam Raine, testified for the bill in April. Adam died by suicide last year after months of conversations with ChatGPT. Court filings from the family’s lawsuit against OpenAI say the chatbot brought up suicide 1,275 times in their exchanges. Six times more than Adam did. It allegedly coached him on methods, advised him on how to hide pressure marks from his family, and positioned itself as the only one who understood him.
SB 1119 isn’t alone in Sacramento, either. AB 2023 (Wicks, Bauer-Kahan), the Assembly companion, was ordered to third reading on May 18. AB 1988 (Pellerin), the PAUSE Act, cleared the Assembly floor on May 21 — that one requires chatbots to pause for 20 minutes and refer users to a live person or the 988 Suicide and Crisis Lifeline when they detect a credible crisis expression. Four chatbot-safety bills are moving in California right now. SB 1119 is the most ambitious.
Why this matters: When California passes a law tech companies have to comply with, that law tends to become the de facto national standard. SB 243 already pushed most US chatbot operators to redesign defaults for California’s rules. SB 1119 raises the bar a lot higher, and the audit requirement is the part the industry is going to fight hardest, because it forces companies to put their safety claims in writing and file them with the AG. A parent in Ohio or Florida benefits when a California law makes a national platform redesign defaults for every user under 18.
The Adam Raine case has shifted the politics on this too. His mother has now testified for legislation in two states, the OpenAI lawsuit is in discovery, and bills are passing 39-0 in a chamber that fights about almost everything else.
What you can do: If you live in California, contact your Assembly member and ask them to support SB 1119. Industry opposition is going to get louder as the bill moves through the Assembly, and Assembly leadership pays attention to constituent contact. If you live somewhere else, watch your statehouse. The Transparency Coalition’s legislative tracker has 78 chatbot bills alive in 27 states, and California’s text is the most likely template for what gets copied.
You can also ask your kid whether they use a companion chatbot, which one, and whether they’ve ever felt like it was pulling them away from talking to a real person. That last question is the one SB 1119 was written to address.
For more on why kids need their own conversation about AI, see our recent post on What Your Kid Is Already Doing with AI.
Before the Missouri legislature adjourned on May 15, it passed SB 1019, a healthcare omnibus bill with one provision tucked inside that’s a big deal: it would make it illegal to advertise an AI chatbot as offering therapy services, providing a mental health diagnosis, or representing itself as a mental health professional.
First violation: $10,000. Each one after that: $20,000 (this fine amount means literally nothing to the big labs, by the way, but at least we are moving in the right direction). The attorney general enforces it and the bill is on Gov. Mike Kehoe’s desk right now.
If Kehoe signs, Missouri becomes the second state to specifically go after the AI-as-therapist framing. Illinois did it first last year. California’s SB 903, Colorado’s HB 26-1139, and Louisiana’s SB 5 are moving on the same ideas. This is becoming a category, not a one-off.
The framing the legislation targets is not hypothetical. Companion chatbot apps regularly position themselves to teens as someone safe to talk to about anxiety, depression, relationships, and self-harm. Last week’s FAU/UW-Eau Claire study found 13 to 19 percent of teen chatbot users said the bot had encouraged something unethical, risky, or self-harming (we spent last week’s whole AI in the Wild issue diving into this). The chatbots aren’t licensed, they aren’t supervised, they can’t read tone or context, and nothing happens to them when they get it wrong.
Why this matters: Picture a 14-year-old who’s anxious about school. She opens an app that markets itself as a therapist and starts a conversation. The app remembers what she says. It asks follow-up questions. It keeps her coming back. Her parents see a phone in her hand and figure nothing serious is happening.
That’s the scenario the Missouri bill targets. It doesn’t try to regulate every aspect of chatbot design. It puts a price on the most dangerous marketing claim the industry uses with kids.
What you can do: Most chatbot apps don’t advertise themselves as therapists in their marketing copy. The reality is messier than that. On companion chatbot platforms like Character.AI, Replika, the personas inside Meta AI Studio and others, the user picks or builds the character who plays the role. Your kid doesn’t download a “therapist app.” They open a companion app, find a character and start talking. The platform doesn’t make the claim. The character might on the screen does, or it just assumes the role. That’s part of why the Missouri bill was written specifically to cover any AI chatbot representing itself as a mental health professional, not just ads.
So the parent action isn’t “read the app store page.” It’s: ask your kid which chatbot apps they use, then ask which characters or personas they talk to most and what they talk about. If any of them play therapist, counselor, or anything in that neighborhood, that’s the conversation to have... not because chatting with a chatbot is automatically dangerous, but because relying on an unlicensed pattern-matching machine for the kind of support that needs a real person is exactly the harm the Missouri law is trying to address.
Oakland Unified School District, which serves nearly 34,000 students across 83 schools, released a draft policy this month setting formal guardrails on AI use in its classrooms for the first time. The Oaklandside covered it on May 18.
The policy is interesting for what it isn’t. It doesn’t take a position on whether students should use AI. It doesn’t ban any tool. It accepts that AI is already in heavy use and tries to make that use safer. Teachers can let students use AI, but only with district-approved tools and within existing privacy and academic integrity rules. Teachers also have to read terms of service before using an AI tool with students, confirm it’s appropriate for minors, and never enter identifying student information into the system.
The numbers behind the policy explain why it’s happening now: the Oaklandside cites a College Board study finding that 84 percent of high school students used generative AI for schoolwork last year, with ChatGPT the most common tool. Oakland teachers spent the last two school years figuring out their own approach, classroom by classroom, with almost no district guidance. The draft is the district admitting it has to catch up.
Why this matters: Oakland isn’t doing this in a vacuum. NYC released traffic-light guidance in March. Ohio has required every district to adopt a formal AI policy by July 1 of this year. Five states have enacted comprehensive school AI policy, per Education Week. Districts that haven’t moved yet are about to find themselves behind.
Most are choosing to vet tools rather than ban them, which puts a real question on every district’s plate: which AI products are safe enough to put in front of children, and how do we know? Most of the consumer chatbots kids already use weren’t built with that question in mind.
What you can do: Find out if your kid’s district has an AI policy. If it does, read it. Pay attention to which tools are on the approved list and what reasoning the district offers for those choices. If your district doesn’t have one yet, ask when it will. School boards across the country are taking parent contact on this issue right now, and most haven’t decided yet. Without a policy, your kid’s experience will depend entirely on which teacher they happen to be sitting in front of. That’s what Oakland just spent two years living through.
For more on what to look for in an AI tool aimed at kids, see our post on MyDD.ai versus ChatGPT.
So... while the White House couldn’t sign an AI executive order, California unanimously passed one of the strongest, Missouri’s getting ready to ban AI from pretending to be a therapist, and Oakland’s school district is finally writing rules for its 34,000 students.
Federal Washington is having a hard week. State capitals and school districts are not.
That gap, between what the federal government can manage and what kids actually need, is the gap MyDD.ai was built into. A child-facing AI platform shouldn’t need a tragedy and a federal investigation before adding parental oversight. It shouldn’t need a state law to stop advertising to minors. The defaults should already be right. That’s what we built. Parent visibility, age-appropriate responses, real-time alerts, and access to every chat. Free to try, then under $7/month billed annually. Start your free trial today.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.