How Developers Are Making AI Your Kid’s Third Parent In The Classroom

Under Roman law a father held a legal power called patria potestas, or “total ownership,” of his children. He could sell them, deny them property, or abandon a newborn on a hillside. The child was not a person but property under the law. What a surprise then that the so-called “paternalistic” Apostle Paul upended five centuries of that system in a single verse when he wrote “Fathers do not exasperate your children; instead bring them up in the training and instruction of the Lord” (Ephesians 6:4). Roman law already demanded obedience to the father under pater familias. So Paul’s revolutionary challenge to the system was not to challenge obedience, but rather to tell the man holding absolute power he had a duty to the best interests of the child rather than himself.

Paul’s words to the Ephesians shaped Western family law for two millennia, including modern American case law (see Pierce v. Society of Sisters, 1925; Wisconsin v. Yoder, 1972). But today a different authority has moved into the space between parent and child; not a patriarch but an “aithority” — an algorithm built by the largest technology corporations on earth and dropped into American classrooms through a partnership with the teachers unions. Nobody sent a permission slip home.

The scale of “the aithority” in schools is already exasperating. In late 2025, Google announced its Gemini AI education tools had reached more than 10 million students across more than 1,000 U.S. institutions. The company rolled out more than 150 new AI features in a single year, trained more than 1 million educators for free, and embedded AI tutoring modules directly into Google Classroom. Separately, Google invested $1 billion in college-level AI integration. In June 2025 the American Federation of Teachers (AFT), the second-largest teachers union in the country, announced a partnership with OpenAI, Microsoft, and Anthropic to accelerate AI adoption in classrooms nationwide. That deal was negotiated between union leadership and three of the most powerful AI companies on earth. Parents were not at the table.

AI’s Values

In AI systems, “features” are not merely technical upgrades. They are a moral takeover, and the people building the technology know it.  

In January, Anthropic CEO Dario Amodei published a 20,000-word essay called “The Adolescence of Technology,” in which he called AI such a “glittering prize” that it is “very difficult for human civilization to impose any restraints on it at all.” He described Anthropic’s approach to controlling its Claude AI through what the company calls Constitutional AI, where developers write a “central document of values” the model reads to form its identity. And if your values are not aligned with pater artificialis Siliconus, Amodei boasts “we can also selectively activate features in a way that alters behavior.” He says the goal is to teach the model “a concrete archetype of what it means to be a good AI” and compared the process to a “letter from a deceased parent sealed until adulthood.”

Read that again. The developers are the parents. The AI models are the adolescents. The actual parents of actual children are excluded from the lineage entirely. This is the new Silicon pater familias. A small group of engineers in San Francisco writes a constitution that shapes how an AI thinks about morality, identity, and the good life. That AI then sits with your child several hours a day in a public-school classroom. The child did not choose it. You did not approve it. But the developers decided what “best interests” and values the model would carry into that conversation and they did so using a framework Amodei himself compares to parenting. 

But to be fair, Amodei is not the only AI morality engineer in the new empire. In 2025 the teacher unions’ other darling, OpenAI’s CEO Sam Altman, basically called ChatGPT a third parent on The Late Show with Jimmy Fallon, admitting he relies on it to manage the anxieties of raising his own surrogate child, saying “I cannot imagine having gone through figuring out how to raise a newborn without ChatGPT.” In contrast to Altman’s idealized helpful third parent, the supposed “Godfather of AI” and Nobel Laureate Geoffrey Hinton warns that with the coming model upgrades, humans will soon be “like three-year-olds” compared to AI, noting only one biological example of a less intelligent being successfully controlling a more intelligent one: a baby and its mother.

This is not the first time a “best interests” framework has been used to justify displacing parental authority at scale. The United Nations Convention on the Rights of the Child (UNCRC) established a model where state-defined interests could override the family unit. Conservative critics from the Heritage Foundation to the Home School Legal Defense Association warned for decades that the UNCRC was an effort to remove children from their religious and family contexts. The United States never ratified it. Now Silicon Valley has effectively digitized the same ambition. The AI constitution acts as a new kind of treaty, one that bypasses local communities and parental sovereignty entirely and that no legislature voted on.

AI at School

A 2025 survey by the Center for Democracy and Technology found that 85 percent of teachers and 86 percent of students used AI during the 2024-25 school year. Only half of those teachers had received even one training session on using AI systems.

Georgetown professor Meg Leta Jones documented for the Institute for Family Studies what happens when a parent tries to opt out. When she attempted to remove her own children from school-issued devices, administrators told her technology was embedded in the curriculum and could not be separated. Under current FTC guidance, AI companies can simply assume the school has obtained parental consent. Jones found that in practice “most parents have absolutely no idea what applications or technologies their children are using in school.” But your child’s third parent AI surely knows that it is exasperating your child, right?

Wrong. The results are already showing. A November 2025 CalMatters investigation found that after Google deployed its AI-powered Lens tool on student Chromebooks in Los Angeles, one teacher watched his students’ grades spike overnight. The students were not learning more. The AI was doing their work. “Teachers and school leaders spend countless hours considering each detail of the learning experience,” the teacher said, “then Google totally undermines it with the click of a button.” Google quietly disabled the feature after the story ran.

But at least the people building these tools seem to understand the problem when their own families are involved. A 2011 New York Times investigation found that the chief technology officer of eBay and employees of Google, Apple, Yahoo, and Hewlett-Packard were sending their kids to the Waldorf School of the Peninsula in Los Altos, California. No computers. No screens. The teaching tools were pens, paper, knitting needles, and mud. Bill Gates and Steve Jobs also raised their children in low-tech households. By 2019, the Waldorf model was operating in more than 1,000 schools across 91 countries. The executives who design the AI being lobbied into your child’s public school pay premium tuition to keep this third parent away from their own kids.

Congressional and Parental Action

There is movement in Washington. Sen. Bill Cassidy’s LIFE with AI Act aims to strengthen student privacy protections and require parental notification before schools use AI tools. Sens. Josh Hawley and Richard Blumenthal have introduced the bipartisan GUARD Act to require age verification on AI chatbots targeting minors. In 2025, an exasperating 53 bills on AI in education were proposed across 21 states. Four became law. Former Surgeon General Vivek Murthy has recently warned that AI risks becoming a “substitute” for human care rather than a supplement, cautioning that “an AI tool … cannot sit with someone through their pain and empathize.”

The real leverage is local, and parents must start using their leverage now. Before any AI tool enters a child’s classroom, the school should be able to answer five questions from parents: 

• Does this software bypass my ability to review what my child is taught? 
• What behavioral hooks are built in to keep my child engaged?
• Whose values trained the model? 
• Does the tool require my child to think or does it hand over a finished product? 
• Can I see every prompt and every response?

If the school avoids answering those questions, the issue is no longer about technology. It is about who holds authority over the formation of a child’s mind.

Amodei is right that AI is like an “adolescent” almost impossible to impose restraints on. Perhaps he and his peers are missing the restraint that matters most, which will not be found in the code, but in what Paul articulated 2,000 years ago in the New Testament. The person holding power over a child has a duty to that child. Not to shareholders. Not to a union contract. Not to a constitution written in a San Francisco office park. When a handful of developers write a values document they compare to a parental letter and then embed it in the classrooms across America, they have assumed the old authority of the Roman father. It is time parents took their authority back before the Silicon Pater Familia exasperates our children.