California AI Safety Debate: Children at Risk or Already Harmed?

A new opinion piece in the Marin Independent Journal challenges the prevailing narrative around artificial intelligence and child safety. The article, distributed by CalMatters.org, argues that the question isn't whether AI will harm children someday, but whose children are already being harmed.

Author Sasha Costanza-Chock, a faculty associate at the Harvard Berkman Klein Center for Internet & Society, critiques a documentary film that interviews AI executives including Sam Altman of OpenAI, Demis Hassabis of Google DeepMind, and Dario and Daniela Amodei of Anthropic. The film reportedly asks whether it's a good time to bring a child into the world.

Costanza-Chock's central argument cuts through the speculative doom-and-gloom framing. According to the piece, Tristan Harris, co-founder of the Center for Humane Technology, tells the filmmaker: "I know people who work on AI risk who don't expect their children to make it to high school." The article notes this line illustrates a core problem: families of elementary school students recently killed in Iran already know their children won't make it to high school.

The piece documents specific AI-enabled harms that the documentary allegedly ignores. On the opening day of the U.S.-Israeli war on Iran, multiple Tomahawk cruise missiles struck the Shajareh Tayyebeh elementary school. At least 168 people were killed, the majority children, according to Amnesty International. U.S. Central Command used Palantir's Maven Smart System for target identification throughout the campaign.

Anthropic's flagship AI, Claude, is integrated with Palantir's systems and was reportedly used in Iran, as well as in the U.S. military operation to capture Venezuelan President Nicolás Maduro. In July 2025, the Pentagon awarded contracts of up to $200 million each to Anthropic, OpenAI, Google, and xAI.

These companies building AI "kill chains" are also causing documented harm within California itself. A campaign by Purge Palantir highlights how Palantir powers the U.S. Immigration and Customs Enforcement deportation machine targeting immigrant communities. Massive new data centers are draining California's water and straining its energy grid during a climate crisis. AI tenant screening algorithms are driving up rents and pushing people out of their homes.

None of this appears in the film. The AI doomers, utopians and CEOs never mention existing AI harms. Consequently, the film ignores the myriad ways communities are already holding AI companies accountable. Resistance to harmful AI systems is real and growing in California.

The Stop LAPD Spying Coalition has led the fight against predictive policing and won, forcing the Los Angeles Police Department to shut down both Operation LASER, built on the Palantir platform, and the PredPol program that uses AI to target Black and brown neighborhoods for extreme policing. The Writers Guild of America went on strike and won groundbreaking protections against the use of AI to replace creative workers.

No Tech for Apartheid, led by Google and Amazon workers in Silicon Valley, has built awareness of tech companies' military contracts with the Israeli Defense Forces. In September 2025, after sustained worker pressure under No Azure for Apartheid, Microsoft blocked Israel from using its cloud and AI services for mass surveillance of Palestinians.

In Monterey Park, residents blocked a massive AI data center and organized to put a permanent ban on the June 2026 ballot. This is the kind of grassroots action that actually changes things (unlike watching a documentary and feeling vaguely concerned).

Meanwhile, California is advancing its own legislative approach to AI safety for minors. A March 2026 Marin Independent Journal piece details the Parents and Kids Safe AI Act, which OpenAI partnered with Common Sense Media to draft. The proposal would require AI systems that simulate conversation to use privacy-preserving age estimation so child-protective settings kick in for users under 18.

The legislation would empower parents by requiring easy-to-use parental controls and implement stronger protections for children under 13. It would include safeguards against manipulative designs like emotional dependency or simulated romantic relationships, and clear crisis-response protocols for self-harm risks. It also calls for independent child-safety audits with accountability through enforcement.

At the federal level, the Senate Judiciary Committee unanimously approved the GUARD Act in April 2026. The bill would make it a crime to knowingly provide a chatbot that might encourage minors into sexually explicit behavior or suicide. It would require chatbot providers to verify ages using government-issued identification or "any other commercially reasonable method."

The bill would also require that platforms limit their data collection to only what is minimally necessary and protect the data from unauthorized access, including by using industry-standard encryption protocols. The Hawley bill would impose a penalty of $100,000 for offering a chatbot that encourages minors to engage in sexually explicit behavior or physical violence.

Both bills are opposed by the technology industry. Amy Bos, vice president of government affairs for industry group NetChoice, called the bill an "overinclusive, blunt mechanism." NetChoice has sued to stop age-verification laws in states around the country based on limits to free speech and has been successful in some cases, at least temporarily.

The physical reality of these systems matters. When a child interacts with an AI chatbot, they're not just exchanging text on a screen. They're building habits, forming dependencies, and potentially being manipulated by algorithms designed to maximize engagement. The friction of parental controls, the delay of age verification, the texture of these interfaces—these aren't abstract concerns. They're the actual experience of using the technology.

Costanza-Chock's piece ends with a pointed question. In the film, the filmmaker wants to know if it's a good time to bring a child into the world. The mothers of Minab in Iran want to know who will be held accountable for the AI-supported mass murder of their children. The question is not whether AI will harm "our" children someday. The question is whose children are already being harmed. And will we demand accountability?

Whether California's new legislation actually protects kids or just creates compliance paperwork for tech companies remains to be seen. The industry has a long track record of treating regulation as a box-checking exercise rather than a genuine safety imperative.

And honestly, if the solution to AI safety is just more age verification and parental controls, we're solving the wrong problem. The real issue is whether we should be building systems that can manipulate children in the first place. But that's a conversation nobody seems eager to have.

Read the full Marin Independent Journal article here.

The real test isn't whether California passes laws. It's whether those laws actually change what happens when a teenager opens a chatbot app at 2 a.m. on a school night. Most of us won't know the answer until it's too late.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn