The Age of AI is upon us, for better or worse. Those who work in the early care and education space are already trying to find the right balance between caution and enthusiasm. Can this space, so long considered “behind the times” relative to other sectors, harness this new technology for good while it’s still fresh? Or do we owe it to families and children to take things slow and proceed with the utmost caution and sensitivity?   

This year marked a key turning point in public sentiment and government posture on this topic. We noticed a significant increase in the mention of (and expectation for) AI-powered functionality in procurement processes, including a subtle shift in language from an emphasis on controls and oversight to efficiency and innovation. While there seems to be general consensus around the need for clear boundaries and humans in the loop, many agencies and organizations seem to be embracing the idea that AI can be a solution for longstanding challenges in this space. 

We’ll continue to monitor how and where those solutions are being tested. In the meantime, let’s look at the other side of the coin—what are some of the negative AI trends in ECE that need to be addressed sooner than later?

1. As an Instructional Resource

Seriously, don’t fall for the edtech sales pitch. The public release of ChatGPT in November, 2022 kicked off an explosion of desperate market pivots, “educational tools” that were little more than shiny wrappers on top of LLMs, and “personalized learning” advances that claimed to engage students by tailoring their instruction to their interests. 

As someone who worked in the field at that time, I can confidently tell you that NONE OF THIS was evidence-based or grounded in any real-world understanding of how we learn. It was all a marketing ploy aimed at either rushing to cash in on the AI gold rush or desperately trying to salvage overvalued software companies in the lead up to the 2024 ESSER funding cliff.

Zero to Three does a great job of summarizing the why behind not using AI with our youngest learners in their excellent resource, AI and Early Childhood Development: Frequently Asked Questions: 

For babies and toddlers, an engaged and attentive adult is more important than any piece of technology. AI tools should never replace all the amazing benefits of face-to-face interaction. Playing and reading together, talking with your little one, soothing and comforting, following their lead, and responding to their cues are exactly the kinds of interactions that young children need to thrive.

While some of the existing literature is more open to the possibility that there may be "non-detrimental" ways to introduce AI as students approach kindergarten, none of it can point to consistent positive outcomes. Even applications of the technology for older children that show hints of promise in studies can easily be replaced by significantly more valuable human interactions. Per a 2026 study from Brookings, "At this point in its trajectory, the risks of utilizing generative AI in children's education overshadow its benefits." 

It says a lot when the only people giving ringing endorsements for educational applications of AI are those who are trying to sell us something. 

2. As a Development Team

As more states and communities look to modernize their ECE systems and replace outdated, legacy builds, new software companies are popping out of the woodwork to seize on the market opportunity. An alarming number are relying heavily on vibe coding in place of (more expensive) software and QA engineers, product teams, accessibility specialists, and UX designers. 

Red Hat summarizes the issue about as concisely as possible: “vibe coding is simultaneously the most exciting and most dangerous development practice to emerge in years.” The ability to prompt code and ship projects with almost none of the knowledge or time overhead traditionally required has already brought to life thousands of ideas that would otherwise have died on the vine. But in the case of public sector software, those benefits are rapidly outweighed by the drawbacks, including: 

• Security vulnerabilities: AI models are trained on public code, much of which features practices that don’t stand up to rigorous security standards. Private information is included in source code, apps become vulnerable to SQL injection, and the same AI used to create can also be used to find backdoors and weak links. 
• Tech debt: AI is notorious for coding to the easiest, most common use cases. It won’t anticipate edge cases, won’t integrate various modules into one cohesive architecture, and won’t leave space for scalability. Mountains of tech debt have crippled legacy systems—let’s not make the same mistake with our next generation infrastructure. 
• Process-centered vs. human-centered development: As much as we might want it to, AI cannot think about or interact with technology in the same way a human being does. Yes; it can make things that (mostly) work, but it can’t conduct and interpret user research, apply years of design experience, or understand the cultural differences that affect equitable access up and down the ECE system. Specialized human expertise will remain a critical piece of the puzzle. 

Vibe coding is here to stay. It’s just too valuable a piece of an engineer’s toolkit anymore to not be deeply integrated into the development stack. The problem lies in vendors trying to replace, rather than supplement, their development teams. Without humans (and their specialized areas of expertise) actively in the loop at every step, vibe coded public sector software is simply not safe or sustainable. 

3. As an Unfettered Family Communication Tool

One of the most popular use cases for AI today lies in helping people formulate their thoughts into coherent written communications. Educators of all age groups have found the technology especially helpful for family engagement, including weekly updates, monthly newsletters, and reminders for upcoming events or classroom expectations. The efficiency is real, adding up to hours of non-instructional time saved and a significant reduction in stress for those who aren’t as comfortable with their writing. 

That said, there still needs to be boundaries between which communications are delegated to a teacher's LLM of choice, and which still need to come from them directly. Progress notes, incident explanations, and sensitive messages about child development or behaviors require heavy human involvement. It’s not necessarily the content of these communications that can cause the most problems, it’s the tone, accuracy, and personal touch that is expected of a professional educator who knows the child and has the context.

In these cases, if you’re going to use AI at all, consider using it to develop an outline or offer suggestions. Anything more, and you run the risk of losing the trust of your families, a situation no educator ever wants to find themselves in.

4. As a Decision Maker

Thankfully, a growing web of legislation at the federal and state levels is aimed at curtailing this issue before it becomes too prevalent. Even so, as AI continues to proliferate, it will be incumbent upon us to not lose sight of the absolute necessity of human-in-the-loop workflows for any use case in which AI would otherwise be used to rank children, prioritize waitlists, determine eligibility, or otherwise automate high-stakes decisions. 

To get to the root of this concern, one must have at least a high-level understanding of AI bias, which is the “systematic and unfair discrimination in the outputs of an artificial intelligence system due to biased data, algorithms, or assumptions.” (AI Bias: 16 Real AI Bias Examples & Mitigation Guide, Crescendo)

This isn’t some made-up, fear-mongering, hypothetical issue. We have seen documented examples of AI bias across industries, perpetrated in many cases by the corporations we interact with every day. Some of the more popular examples include: 

• Amazon’s failed launch of an AI recruiting tool that demonstrated bias against women because the hiring data it was trained on was heavily skewed toward male candidates in a traditionally male-dominated field.
• Wisconsin’s Dropout Early Warning System generated false alarms about Black students at a 42% higher rate than their white counterparts because its “risk scores” mapped historical performance of racial subgroups to predict likelihood of graduation.
• A 2019 study showed that one of the most widely used algorithms in the United States health care system inequitably identifies who needs care because it “uses health costs as a proxy for health needs.” Black patients historically have less access to healthcare, thus less money is spent on them, and the algorithms see that as an indicator of lower need.

AI systems are biased because humans are biased, and the fact that they’re trained on historical data means they are more likely to perpetuate than alleviate systemic bias and discrimination. 

5. Without Transparency or Consent

As ethical concerns, data privacy awareness, and ongoing legal challenges raise mainstream visibility into what’s going on behind the curtain, we’ll almost certainly be seeing a lot more talk about consent and right of refusal. It’s already illegal under FERPA to dump identifiable and protected student data into public models, but it’s still happening every day. 

Educators must be transparent with families about how AI tools are being used and what information they are being fed. This is a lighter lift in school districts, where AI policies are increasingly likely to be built into student handbooks and start-of-year communications, but child care centers, family child care providers, non-school district preschool programs, and others will eventually need to follow suit. 

The vendors who are building our ECE technology infrastructure must also be cautious about overly relying on AI without alternatives. It’s not hard to imagine a future in which organizations or even individual users have the right to refuse the use of AI tools. This has most recently been proposed by a group of nearly 100 high school students from all 50 states in the Students First Act (written for K-12, but certainly applicable to ECE). Consent management will also be of the utmost importance, as vendors will likely need a clear audit trail of permissions for data sharing and inclusion in any AI-powered functionality. 

These concerns add overhead to development efforts and they introduce technical challenges. Realistically, they aren't likely to be prioritized until required by law or contract. That said, these things are always a lot harder to bolt on later. The writing is already on the wall—transparency and consent won't be optional forever.

When in Doubt, Vet it Out

The growing optimism around the safe, structured, and intentional use of AI in government sectors points to a future in which the technology is no longer “nice to have,” but an essential, embedded component of the larger ecosystem. This shift brings the promise of more productivity, more consistency, and potentially a better-functioning system from top to bottom. It’s also following the well-worn path of dozens of “gold rush” inflection points throughout history that spawned equal parts innovation and exploitation. 

How can we keep from falling prey to the latter category? It's the old adage of “trust, but verify,” though opinions may be mixed on how much trust we should freely give at this point. Generally speaking, AI works best when it’s accompanied by: 

• Oversight from those who are deeply knowledgeable in how it works, the context in which it is being implemented, and the implications of mistakes or misuse. 
• Human-in-the-loop feedback models whenever possible. Let AI do the dirty work or offer recommendations, but leave the critical thinking to people.
• End-to-end transparency. Vendors and the organizations using their products must both be very clear about where, when, why, and how. The stakes are too high for us to trust in good intentions. 

How is AI making your life easier today? Where have you seen it fall short? Send us a message at childcarematters@getbridgecare.com—we'd love to incorporate your experiences into future articles on this crucial topic.