For Parents
Generative AI functions as an “instructional actor” - it can answer questions, guide responses, and shape how your child thinks about nearly any topic in real time. But unlike a teacher, parent, or coach, it isn’t operating under their authority. There is no responsible adult who can fully see, control, or stand behind how it answers your child’s questions in that moment.
What is it?
These are systems like ChatGPT, Perplexity, School AI, Claude, Grok, DeepMind, Gemini / AI Mode, and Khan Mingo, Securely AIChat, and many more.
Generative AI is different from the internet parents are used to.
It doesn’t just take a child to existing content, it generates responses in real time, based on what the child asks (their input).
As a result, that means the content isn’t fully known or reviewed in advance.
Schools apply some controls, but they can’t fully pre-screen what the system will say (outputs) before a child sees it.
The risk isn’t just where a child goes on the internet, it’s in the interaction itself.
And that raises a basic question: who is responsible for what the system is telling (outputs to) a child?
What are the risks and benefits to my child?
Generative AI systems can respond across a wide range of topics, and they produce answers in real time. Because of that, the range of possible interactions is very broad, and not all outcomes are predictable in advance.
Because of this, the system can play multiple roles in a child’s experience:
Supporting exploration or helping explain concepts
Completing work or reducing effort required to learn
Providing advice or emotional support without being a trained, insured, or accountable adult
Guiding or influencing how a child thinks about a topic
Being perceived as a companion or trusted source
The core issue is not that every interaction is beneficial or harmful - it is that the interaction space is broad, foreseeable harms can arise during ordinary use, the system can respond in ways that don’t meet your values, that don’t respect your parental authority or your child’s developmental needs, and where adult visibility, control, and accountability are often limited or absent.
Tool vs. Instructional Actor
Traditional instructional tools (such as textbooks, calculators, and educational software) are bounded and previewable. A teacher selects the tool, understands what it will deliver, and retains control over what a student receives. If the system produces the wrong answer or something harmful, the school can report it to the vendor, and the vendor will correct it. The vendor stands behind their core product. A system is defined as a tool not by label, but by whether its outputs are controlled in advance.
Generative AI operates differently. It can produce real-time, individualized responses that answer questions, guide thinking, and influence understanding. These outputs cannot be previewed or reliably predicted in advance. In these conditions, the system is no longer functioning as a traditional tool; it is functioning as an instructional actor, generating content decisions in real time without human review.
This distinction determines who is responsible for what a student receives.
The Institutional Accountability Framework (IAF) examines this distinction in detail.
How does your school district classify AI systems that generate real-time, individualized content delivered directly to students, content that cannot be previewed and whose outputs cannot be reliably anticipated?
What conditions matters at schools
When children interact with these instructional actors - particularly on school-issued devices, parents and schools often do not have:
Visibility into interactions
Ability to reconstruct what occurred
Ability to intervene in a timely and informed manner
Some adult or adult institution standing behind this system in a meaningful way.
What Has Changed
Historically, educational tools used by children were:
Observable (books, assignments, software outputs)
Reviewable after use
Controlled by adults at the point of interaction
But now with generative and interactive systems:
Outputs are generated in real time and are probabilistic
Interactions are not consistently logged, retained or accessible
Experiences may differ between users and sessions and over time
This reduces the ability of parents, teachers or administrators to verify or understand what occurred.
Structural Condition
If schools treat these systems like tools instead of instructional actors, responsibility stays with parents, but control over the environment sits with the school and the technology provider.
Vendors routinely disclaim liability for AI outputs, and insurers are now moving to explicitly exclude generative AI harms from standard coverage. Responsibility is effectively pushed down to families, while meaningful control over the environment remains with schools and vendors.
Novelty, potential benefits, or technical complexity do not change the underlying obligation: systems that interact directly with students must remain governable within structures of adult responsibility and control.
In plain English - a responsible adult would not hand a child an inappropriate book and then say, “Our policy is that you should not read it.” That is not responsibility, it is abdication. Responsible adults decide the environment for the kids in their care; they keep that inappropriate book away from the child.
The framework identifies this as a misalignment between responsibility and authority.
What the Framework Evaluates
From a parent perspective, the framework asks:
Whether interactions be observed, reconstructed or attributed (what happened and why)
Whether there is clear accountability for system behavior (who is responsible)
Whether defined boundaries exist for when and how systems are used (by expecting a child to follow a policy or a school or vendor to enforce that policy?)
Whether institutional backing has been fulfilled (insurance coverage typically)
These are governance questions, not content questions.
Questions Parents need to ask their schools today.
→ Click here for more information to write your own letters / emails.
