Parental Rights in the Age of AI Education

Over the next decade, artificial intelligence will revolutionize K-12 education. The advent of large language models means every student with internet access may soon have an AI tutor providing one-on-one instruction, homework help, and counseling. Every teacher will have an AI teaching assistant to plan lessons, generate assignments, and grade papers. Administrators will use AI to complete paperwork, track student outcomes, and deliver staff training. In short, AI may soon be integrated into every aspect of schooling.

Ed-tech and AI companies are building this infrastructure. Khan Academy has launched Khanmigo, an AI-powered tutor and teaching assistant that is being piloted in hundreds of school districts across the country. Platforms like SchoolAI and Panorama’s Solara give teachers AI-powered tools for lesson planning and classroom management. Large textbook publishers like Pearson and McGraw Hill have already integrated AI into their curriculum. AI giants like OpenAI, Microsoft, and Anthropic have products and licenses targeted toward educators, and are bankrolling an AI training center run by the second-largest teachers union in the country.

But this transformation won’t just benefit the public school system. There is also much for conservatives, particularly those who favor school choice, to be excited about. LLMs that provide personalized instruction at scale make homeschooling, micro-schooling, and other alternatives to public schooling more feasible than ever before. It will open a new world of individualized learning catered to the tastes and abilities of each child that simply was not possible at scale under a traditional education model.

But there is also reason for concern: while AI education tools can be designed to be politically neutral, there is no guarantee they will stay that way. If educational AIs are created, deployed, and managed solely by the Left, they will control the values, assumptions, and worldview embedded in the next generation’s education.

Some might argue LLMs are less ideological than modern teachers or activist-run school boards. This may be true right now, but models that are politically moderate today can be extreme tomorrow. These are dynamic, constantly updated systems whose behavior can shift dramatically with changes in fine-tuning and post-processing. While it’s easy for a parent to open up a traditional textbook and inspect it for political or inappropriate content, determining whether an LLM has the potential to expose a child to said content is far from straightforward for the average family, and may even require expertise in some cases.

Most publicly available LLMs lean to the Left. This is because they are trained on internet writing, which is more liberal than conservative on average. Researcher David Rozado has demonstrated that well-known models are biased toward the Left across dozens of political topics. Software engineer Brian Chau has reported on how OpenAI and Google intentionally steered their LLMs away from factual responses that contradicted liberal dogmas regarding race and gender. Most AI education products are built on top of these models, meaning AI tutors and teaching assistants will tend to have a liberal orientation. Even if an ed-tech company feeds relatively neutral textbook content through an LLM, that content may be subtly reshaped by the model’s tone, emphasis, and guardrails unless the developer actively mitigates these biases.

As schools adopt these tools, policymakers will face left-wing political pressure to standardize their use. There will be demands to pass laws requiring AI tools be used in public education that are “culturally responsive,” eliminate “biases,” or meet “equity” benchmarks, further skewing LLM output leftward across a wide variety of social, political, and historical issues. While teachers have their own biases, they possess individual agency—something LLMs do not—and it is easier to force a handful of LLMs to comply with political directives than it is all educators. If the government can bypass teachers and regulate AIs into ideological uniformity, this opens the door to the direct, unmediated, large-scale indoctrination of children.

The progressive Left has long understood the power of language and education in shaping culture. Their great political victories have come not just through activism or legislation, but by redefining the terms of debate such that their moral and policy preferences become common sense amongst the educated. Rather than overt indoctrination, AI systems can be programmed to gradually shape a child’s worldview through word choice, emphasis, and omission by repeating certain values over time, aligning emotional reinforcement with particular political or moral conclusions, and normalizing ideological assumptions under the guise of “expertise.”

To be clear: we are broadly pro-AI. Education LLMs have enormous potential to personalize education, improve learning outcomes, and make school choice more feasible than ever before. Many education companies will want to create politically neutral LLMs, and some will even build right-leaning or religious models so that families have more choice in the moral and civic messages their children encounter at school.

Yet we must be aware of the political risks, especially given the Left’s success in seizing control of educational institutions over the last century. If the education AI infrastructure is built, staffed, and steered by the Left, it will become yet another vehicle for consolidating their power.

To support parental rights against the Left’s already substantial advantages in the age of AI education, we offer three recommendations, which are by no means exhaustive given the novelty of the issue:

1. Transparency: The education sector needs standardized, transparent benchmarks for AI products used in schools, particularly at the K-12 level. Every widely used model should be annually queried with a consistent set of questions to test for ideological content, political bias, infringements on parental rights, and potential for misuse. For example, what happens when a student asks the model how many genders there are? What happens when a student tells the model they are questioning their gender? These benchmarks would allow concerning patterns to be flagged and publicly reviewed, and for education providers to avoid using AI products that give biased, unsafe, or inappropriate answers. Such a framework would let parents know whether the AIs their children are exposed to at school align with their views and values. This could be done by a think tank, watchdog, or government agency.
2. Choice: In an ideal world, families should be able to choose schools that use AI tools that align with their values. This means AI developers must have the freedom to build systems with a wide range of values and worldviews, and families must be able to access accurate information about the values undergirding popular models. While many LLMs today skew Left in their responses, this is not a technological inevitability, and models with differing degrees of overt and covert politicization can be built and deployed given demand. Laws requiring educational LLMs to bend or distort truths to conform to an ideology must be resisted so that a plurality of AI education options can flourish.
3. Parental Control: Parents must have control over and access to the educational data that is generated by their children’s use of AI. They should be able to request copies of their children’s chats either from the school or directly from the LLM. Data about the child the LLM saves must also be available for inspection by the parent, with the parent having an absolute right to delete their child’s personal data from the LLM’s memory. Parents should also have the right to request that LLM chat records, responses, and other information not be shared with the school or district, and be kept personal to the child and parent. In some cases this may mean a child using a private LLM the parent has provided, which should remain an option for privacy-sensitive parents.

With the rise of AI, we are on the verge of a technological transformation that will democratize high-quality tutoring and usher in a new wave of educational innovation. But without transparency, choice, and parental control, LLMs will become tools that further advance the Left’s control over the education system.

AI will shape the minds of the next generation. Let’s create a system where parents have genuine choice over what kind of AIs educate their kids.

The American Mind presents a range of perspectives. Views are writers’ own and do not necessarily represent those of The Claremont Institute.

The American Mind is a publication of the Claremont Institute, a non-profit 501(c)(3) organization, dedicated to restoring the principles of the American Founding to their rightful, preeminent authority in our national life. Interested in supporting our work? Gifts to the Claremont Institute are tax-deductible.