Virtually every member of Congress wants more action to be taken to protect kids from harm caused by social platforms that incorporate artificial intelligence. Yet the question of whether tech companies must take steps to prevent their products from causing users harm to avoid liability is causing hang-ups that are complicating the potential passage of new laws this month.
On Monday night, the House of Representatives passed the Kids Internet and Digital Safety (KIDS) Act on a bipartisan basis, 267-117. The sweeping children’s safety package combines text from over a dozen other bills. Its provisions, among others, include blocking platforms from using children’s personal data for targeted ads and requiring AI chatbots to provide crisis intervention resources if the child prompts the chatbot about self-harm or suicide.
It also includes a small line saying that the legislation does not “impose a duty of care on a provider of a covered platform.”
Why We Wrote This
Lawmakers are working to pass laws that protect kids online. One route forces Big Tech to change harmful algorithms, making failure to do so illegal. The other relies on voluntary rules that critics deem ineffective.
That detail makes the bill’s passage in the Senate deeply complicated, as the Senate must still approve the bill for it to go on to the president’s desk.
“Duty of care” is a broad term that means a tech company must take reasonable steps to prevent users from being harmed by its products – by not intentionally designing hyperaddictive features, for example – or face liability.
The congressional divide on the bill does not follow simple partisan lines; it is mainly a split between the House and the Senate. Many members of Congress – Republican Sen. Marsha Blackburn of Tennessee and Democratic Sen. Richard Blumenthal of Connecticut among them – say that including duty of care in legislation to protect children online is a necessity. Others won’t tolerate a bill that includes it.
How the debate plays out could define how the government establishes the parameters for AI accountability going forward.
“These kinds of automated tools, they are creating a lot of harm,” says Danielle Citron, a law professor at the University of Virginia who teaches about online privacy. “Right now, we’re on the cusp of figuring out how to have companies internalize those costs.”
Children’s safety online
The degree to which digital platforms can be held liable for harming users, particularly children, is a debate that’s been playing out since even before AI entered the conversation.
In a California lawsuit, a jury found in March that Instagram’s parent company, Meta, through the negligent design of addictive platforms, was 70% responsible for a young woman’s mental health struggles and negative thoughts about her body. In a separate case, a New Mexico jury found Meta liable for consumer protection violations. Meta has said it will appeal both decisions.
Now, the popularization of AI – and the resulting questions around issues such as whether chatbots might encourage young users experiencing a mental health crisis to harm themselves – appears to have given a new push to efforts to pass reforms at the federal level.
In addition to the KIDS Act, which passed the House Monday night, Ms. Blackburn is negotiating with the government to finalize text for a similar AI safety bill that would incorporate duty of care – as well as a controversial provision that would block some state laws on AI to make way for a federal law.
Republican Sen. Ted Cruz of Texas plans to advance a package that includes some similar provisions, as well as a ban on children under 13 using social media. He has indicated he is open to a duty-of-care requirement.
In June, Reuters reported that Meta had lobbied members of Congress to pass provisions to shield it from child-harm lawsuits targeted at online platforms like Instagram. The current version of the KIDS Act in the House of Representatives does not include any such provision.
“AI debates have given a huge boost and brought a lot of attention to a lot of these questions [around duty of care],” says Ashkhen Kazaryan, a senior legal fellow at the Future of Free Speech at Vanderbilt University. (She previously worked on Meta’s content regulation team.)
Questions around duty of care
Eric Goldman, a professor at Santa Clara University School of Law, argues that the duty-of-care requirement is fundamentally ambiguous. Content that might harm one child could be neutral or even beneficial to another, he says.
He cites the LGBTQ+ community as an example. LGBTQ+ children often rely heavily on support groups found on the internet, he says, which they sometimes access anonymously and through which they might share sensitive health information.
In a bill Ms. Blackburn introduced in 2025 – and which she is expected to use as the basis for her negotiations with President Donald Trump – one provision bars companies from designing features that could contribute to people’s “compulsive” use of online platforms.
To Mr. Goldman, this is a perfect example of an ambiguous law that could cause harm despite being designed to prevent it. While some children might be unable to stop using a service because it’s addictive, he says others might use it extensively because it helps them address a problem.
“What’s the line between compulsive usage and people who are finding benefit in a service in a way that’s actually improving their lives?” he asks. “Compulsive usage doesn’t have any definition from a medical or psychological standpoint.”
Professor Citron says duty of care can sometimes be implemented in a way that is too vague. But she believes lawmakers can remedy that by drafting more specific laws. “If lawmakers really got down to it, they can figure [it] out,” she says.
More important, she adds, is making companies accountable for doing things such as creating models they know to be addictive, or designing products where it’s possible for children to come across things like sexually explicit, deepfake videos. She argues neither parents nor children have the tools to safely navigate an increasingly complicated online world. Leaders of companies such as Meta and OpenAI have said their products are not designed to be addictive and that they prioritize safety for younger users.
“Companies aren’t internalizing the costs that they’re externalizing, and kids are the losers,” Professor Citron says.
Some people, such as Republican Sen. Rand Paul of Kentucky, who opposed a duty-of-care provision in a 2024 kids online safety bill, say such provisions can lead to free speech and censorship issues.
Dr. Kazaryan, the legal scholar from Vanderbilt, shares those concerns. Under duty of care, she says, platforms would have to make judgments about issues protected under the First Amendment, such as whether body positivity is a form of healthy self-love or a promoter of obesity.
“You don’t want government deciding what speeches or items of speech are wrong,” she says.
She is also concerned that, if companies fear retaliation, they will overcorrect and in an overabundance of caution eliminate valuable content.