On Wednesday, August 5, the Senate Commerce Committee did something Congress rarely manages: it moved four different AI and kids’ safety bills out of committee in a single afternoon, three of them without a single recorded objection. For an institution that has spent years talking about regulating AI chatbots without actually doing it, that is a genuinely unusual pace. It is also not the full story, because the same afternoon produced a public falling out between the people who normally agree on protecting children online.
What the Committee Actually Voted On
Four separate bills moved forward that day, each with its own sponsor and its own narrow slice of the “AI and kids” problem. Here is what each one does, in plain terms.
| Bill | Sponsors | What it does | Result |
|---|---|---|---|
| KOSA | Blackburn (R-TN), Blumenthal (D-CT) | Requires platforms to limit harms like self-harm content, addictive design, and exploitation for users under 17, with safe defaults | Advanced unanimously, as amended |
| CHATBOT Act | Cruz (R-TX), Schatz (D-HI), Curtis (R-UT), Schiff (D-CA) | Mandates parental consent and monitored “family accounts” for teen chatbot use | Advanced without objection |
| Youth AI Privacy Act | Markey (D-MA) | Limits chatbot data retention and use for minors, bans some targeted ads and training uses | Advanced with a revised 30-day retention cap |
| Children’s AI Toy Safety Act | Duckworth (D-IL) | Directs a federal review of privacy and safety risks in AI-enabled toys marketed to kids | Advanced without objection |
| SCREEN Act | Lee (R-UT), Curtis (R-UT) | Broader age-verification mandate for adult content, unrelated to AI specifically | Failed, 15-13, on a quorum shortfall |
The CHATBOT Act Wants Every Teen Account to Come With a Parent Dashboard
Of the four bills, the CHATBOT Act reaches deepest into how AI products actually work. It requires covered chatbot providers to offer parents a “family account” that includes a full record of a teenager’s conversations and activity, plus tools to monitor and analyze that activity at scale. Providers also have to alert parents if a teen tries to get around the controls.
That is a meaningfully different model from what most platforms offer today, which is usually a parental-controls toggle a family can choose to turn on. Under the CHATBOT Act, the monitoring infrastructure has to exist for every provider, and it has to be presented as part of the consent process before a teenager can use the product at all. Supporters frame this as giving parents a tool they currently lack. Critics point out that once a company is required to keep a permanent, centralized log of a teenager’s private conversations for parental review, that log becomes a target in its own right, for hackers, for identity thieves, and in messy custody disputes where a parental account login is the least of anyone’s problems.
The Youth AI Privacy Act’s Fight Over Memory
The Youth AI Privacy Act had the rockiest path through markup. As introduced, it would have handed the FTC authority to set rules on how long a chatbot could keep using a minor’s personal data to shape future responses. The committee’s substitute stripped that FTC rulemaking power out entirely and replaced it with a flat 30-day default retention limit, extendable only if a parent gives verifiable consent.
Senator Markey defended keeping the cap against an amendment that would have removed it, arguing that without a limit, “the 30-day limit is a reasonable guardrail against effectively having a permanent memory for minors.” The bill also bars companies from using a minor’s chat logs to train models or sell that data to other companies for training purposes, which is one of the few provisions in this package that privacy advocates and industry both find easier to live with than the rest.
KOSA Comes Back With 75 Co-Sponsors
KOSA is not new. The Senate passed an earlier version 91-3 back in 2024, only for it to stall out in the House. This time, Senator Blackburn told the committee the bill now carries 75 Senate co-sponsors, a number large enough to make floor passage look likely if leadership schedules a vote. She used the markup to draw a sharp line against the House’s own version of the bill, which passed earlier this summer, calling it “toothless” and “a pale imitation” of what the Senate is proposing.
That gap between the two chambers is worth watching, because it means even a clean Senate passage does not end this story. Whatever reaches a floor vote will eventually have to be reconciled with whatever the House is willing to accept, and right now those two bills do not look alike.
Why Child Safety Groups Are Calling This a Win
The Electronic Privacy Information Center and more than a dozen coalition partners sent the committee a letter the day before the markup urging support for the Youth AI Privacy Act specifically, citing its limits on manipulative chatbot design, its restrictions on targeted advertising to minors, and its ban on using kids’ data for model training. For groups that have spent the past two years documenting cases where AI chatbots were linked to real harm to minors, including wrongful death lawsuits against AI companies, a unanimous committee vote on four separate bills in one day is the kind of legislative momentum that rarely shows up.
Why Civil Liberties Groups Are Not Cheering
The Electronic Frontier Foundation opposed all four surviving bills, and its objections are less about the goal than the mechanism. Its core argument is that any law requiring extra protections for users under a certain age forces companies to figure out who is under that age in the first place, which means more age verification, not less. EFF has been tracking this dynamic since well before this markup, pointing to how a separate wave of state and federal age-verification laws has already pushed platforms toward collecting IDs or running facial-age-estimation scans on users who would rather not hand either over.
EFF’s second objection is more specific to the CHATBOT Act: mandating a single, federally prescribed monitoring architecture for every teenager, whether they are 13 or 17, takes away the ability of individual families to decide how much independence their own kids get. Its third objection is a First Amendment one, arguing that “safe design” mandates function as speech regulation dressed up as a safety feature, and that similar state-level design codes in California, Texas, and Arkansas have already been blocked in court on exactly that basis.
Silicon Valley Isn’t Thrilled Either, for Different Reasons
Industry pushback came from the Computer and Communications Industry Association, whose members include Google, Meta, and Amazon. In a letter sent ahead of the markup, CCIA said it shares the goal of protecting young users but argued the bills would replace parental discretion with what it called “prescriptive federal mandates” built on vague and inconsistent standards, warning of both constitutional exposure and heavy compliance costs. That complaint lands against a backdrop where AI companies are already facing legal pressure over chatbot safety from another direction entirely. Florida became the first state to sue OpenAI directly over alleged ChatGPT safety failures involving minors, and state attorneys general have shown they are willing to move faster than Congress when they think federal action is too slow.
That puts AI companies in an uncomfortable middle position: too slow for child-safety advocates and state prosecutors, too heavy-handed for civil liberties groups, and now facing a specific federal compliance burden on top of whatever states decide to do on their own.
| Group | Position | Main concern |
|---|---|---|
| EPIC and child-safety coalition | Supports | Wants stronger data and design limits for minors |
| EFF | Opposes | Age verification pressure, First Amendment risk, new honeypots of teen data |
| CCIA (Google, Meta, Amazon) | Opposes | Vague standards, compliance cost, constitutional exposure |
| State attorneys general | Acting independently | Already suing individual AI companies over alleged harm to minors |
What Happens Next
- A full Senate floor vote. Committee passage is not law. All four bills need to be scheduled and passed by the full Senate, and Majority Leader scheduling has stalled similar bills before.
- A House reconciliation fight. The House already passed its own, weaker version of KOSA. Any final bill has to bridge that gap, or the Senate’s stronger language could get watered down in conference.
- A possible SCREEN Act rerun. The bill failed on a quorum issue, not a policy defeat, meaning Chairman Cruz could bring it back for another vote once enough senators are in the room.
- Compliance planning for AI companies. None of this is enforceable yet, but the direction is clear enough that any chatbot provider with a meaningful teen user base should be scoping what a mandatory family-account system and a 30-day data retention default would mean for its product.
None of these bills are law yet, and the distance between “unanimous committee vote” and “signed legislation” has swallowed plenty of past kids’ safety bills before. But the fact that four different measures moved in one sitting, with support that cuts across party lines, suggests Congress is closer to actually regulating how AI chatbots interact with minors than it has been at any point since the technology went mainstream. The disagreement now is not really about whether kids need more protection. It is about whether the fix on the table protects them, or just makes their data easier to find.
