Every major player in the AI wars wants to claim the same role: the good guy. “Trust us,” they say. “We are the responsible ones.”
OpenAI and Anthropic have been jockeying for position in the AI “good guy” standings. At the end of last month, OpenAI announced that it was delaying the release of GPT 6.1 due to security concerns raised by its researchers.
Not to be outdone, Microsoft is entering the race by partnering with the American Federation of Teachers (AFT) to develop and release a “National AI Safety & Privacy Standard” for schools.
Their stated goal is admirable: “legally enforceable protections to ensure the safety of students, families and educators.”
But are we really expected to believe that Microsoft is motivated solely by an altruistic desire to keep students safe when they use AI in school? Or is the company positioning itself as the trustworthy choice for schools preparing to bring AI into the classroom?
As Yogi Berra famously said, “It’s déjà vu all over again.” We have already seen this playbook.
If we have learned anything about Big Tech over the last three decades, it is that investing in K-12 schools is decidedly not about improving learning outcomes. It is about market capture and long-term growth.
Internal Google materials made that strategy startlingly clear. A 2020 presentation stated: “If you get someone on your operating system early, then you get that loyalty early, and potentially for life.” The company's education strategy contemplated a “pipeline of future users.”
And it worked.
School districts have increasingly aligned themselves with specific technology ecosystems, often identifying as “Google districts,” “Microsoft districts,” or “Apple districts.” Once a district has invested in devices, software, cloud services, and teacher training, switching becomes expensive and inconvenient. Every new product that integrates with that existing infrastructure comes with an enormous built-in advantage.
Meanwhile, taxpayers have spent billions putting devices into children's hands and transforming classrooms around digital technology, with remarkably little evidence that any of it actually improved learning.
In fact, the evidence increasingly points in the opposite direction. As Jared Cooney Horvath writes in "The Digital Delusion," “Far from doubling learning in half the time, digital tools appear to be doing the opposite: halving the learning in double the time.”
Let that sink in. Schools undertook an enormous technological experiment on a generation of children. Technology companies gained millions of young users, school districts became increasingly dependent on their products, and taxpayers picked up the bill.
Beyond the hundreds of millions of taxpayer dollars these companies receive, they have made millions more capturing and selling student data to marketers for targeted advertising. In 2014, Google acknowledged that it had scanned students’ e-mails in its education platform for targeted-advertising purposes, and Microsoft is currently facing a proposed class-action lawsuit in Washington State alleging that the company “collects, extracts, and generates vast troves of student data, converts that data into hyper-detailed psychographic profiles of students, and uses that data for its own commercial and other purposes, enriching itself at students’ expense.”
Now we are preparing to do it all over again. Only this time, the product is AI.
It is not difficult to see what Microsoft stands to gain by establishing itself as the “safe” AI company for schools. The company whose AI becomes familiar — and eventually indispensable — to today's students and teachers gains an enviable position in tomorrow's market.
Which is exactly why we should be skeptical of allowing any tech company to define what “safe AI” in schools looks like.
A product can comply with a privacy standard and still undermine learning. It can protect student data while encouraging intellectual dependency. It can boast impressive safety features while displacing the reading, writing, struggling, memorizing, and thinking through which children develop their own intellectual capacities.
That is the danger of a “safe AI” standard; it can easily become a permission slip: This product has been certified safe. Go ahead and deploy it.
The debate over AI in schools cannot begin and end with whether the technology meets a set of safety and privacy standards, especially when the companies that stand to profit from its adoption are helping write those standards.
Before asking how to make AI safe enough for schools, we need to ask a more basic question: How much AI do children need in school in the first place?
And before we accept “more” as the answer, we should ask who stands to benefit from it.
Melissa Henson is the Senior Policy Advisor for Media and Culture for Concerned Women for America, the nation’s public policy women’s organization, dedicated to promoting biblical values and constitutional principles in public policy. On X: @CWforA