In an Oct. 13 webinar, “A National Conversation on AI in Education: Policy, Privacy, and Practice with Randi Weingarten,” hosted by the Campaign for Grade Level Reading, education leaders examined the goals, challenges and implementation of an emerging framework for schools and the use of artificial intelligence (AI). The framework, the National AI Safety & Privacy Standard for Schools, is a partnership between the American Federation of Teachers (AFT), United Federation of Teachers and Microsoft.
AFT President Randi Weingarten introduced the framework, which has 10 essential principles built in. The document is meant to be used as a Memorandum of Agreement when bringing on new education technology (edtech) providers with AI tools:
- No student or educator data for training — ever, with one narrow safety exception: The narrow exception “applies only to processing that is reasonably necessary to detect, prevent, investigate, mitigate, or remediate harm to users or threats to the security or integrity of the AI Provider Educational Products, including self-harm risk, child sexual abuse material, grooming, threats of violence, bullying, malicious activity, security incidents, vulnerabilities, or unauthorized access,” according to the framework.
- Data minimization by design: The AI provider may only collect, process or store data that is strictly necessary to deliver the specific AI provider educational product for which a customer has contracted.
- The customer owns its data — completely: All covered data, including every prompt, response, file, and AI-generated output, belongs exclusively to the customer and its students.
- Human oversight and explainability: AI tools should be used to support educators and students and do not replace human judgement. “No AI system covered by this Standard may make decisions for the EDU Customer’s educators and students, without allowing for meaningful human review or prior approval,” the framework states.
- Real accountability — enforceable, not just promised: The AI provider must pursue or maintain current, independent third-party certifications appropriate to the sensitivity of educational data.
- Security that matches the stakes: Student data is among the most sensitive data that exists and security must reflect that. The document lists minimum security standards.
- Meaningful consent and transparency for families: Requires the AI provider, within 90 days, to develop and make available educational transparency materials regarding its products, including a plain-language guide to what the tool does and a frequently asked questions resource.
- Equity, access and inclusion: AI tools must work fairly for every student, regardless of race, language, disability or economic background. Accessibility features must be built into the product, tools must be tested across diverse student populations, and all features must be offered to all students equitably.
- No feature creep and no vendor lock-in: Educators, not vendors, decide how AI provider products evolve in the classroom.
- Long-term data responsibility: The AI provider’s obligations to protect student data do not end when the contract ends. Some commitments, such as the prohibition of using covered data for training, are permanent.
“The first issue we were trying to solve was how do we not make the mistakes that were made in social media on [student] well-being?” Weingarten said. “And that’s why we worked with these edtech companies. It would be better if that safety standard was based in regulation, based in law, but the absence of the federal government having any interest in this, we felt like we couldn’t wait.
“What should the guardrails look like?” Weingarten continued. “No screens in K-2 and AI-facing instruction only starting in middle school or high school. Working with Microsoft, our insistence was that this had to be real for parents, there had to be a standard of privacy, there had to be real transparency, there had to be agency for parents, and it had to be broad. Microsoft agreed on the age limits and that by Nov. 1, every one of their contracts with school districts would include this standard moving forward.”
Panel discussion
Panelists emphasized the importance of protecting student and staff data, transparency in what an AI company is providing, the importance of the human-centered learning environments and decision-making, and procurement and equity considerations when choosing edtech with AI components.
“What’s helpful about this framework is that it doesn’t bind districts, but it signals what they can expect and what they can demand,” said Eisha Buch, head of teaching and learning at Common Sense Media. “And I think that is essential given the rapid evolution of this. Oftentimes districts don’t have the expertise on their team to do this vetting. So, to have that as the floor, I think, is essential and fundamental.”
Ray Hart, executive director with the Council of the Great City Schools (CGCS), spoke about the importance of keeping data secure. “First and foremost, the safety and security of our students is paramount. And with AI in particular — really understanding what safeguards need to be in place to help educators, and heaven forbid students misuse or other staff misuse AI, understanding what alerts need to be in place and kind of what follow-ups are there.”
Chris Lagoni, National Rural Education Association president-elect, shared how the framework aids rural districts with lean staff and tight budgets. “The National AI Safety Privacy Standards can help as a procurement and decision-making compass. It can also reduce workload of the if, when and how AI is adopted,” he said. “It could be a clear no-go gate for a vendor if their standard baseline requirement doesn’t meet these high standards when it comes to model training, teacher data, breach notice, timelines, human in the loop, things like that. And it also builds community trust. I think it’s going to help every community that values transparency and student safety.”
Panelists talked about the importance of community support through transparency and the need for careful implementation.
“Anything you’re going to implement around these big, important concepts really will rely on local buy-in, local engagement for people who aren’t going to be in the classroom, so parents feel empowered, so parents feel heard, so parents understand,” said Noelle Ellerson Ng, chief advocacy and governance officer with AASA, the School Superintendents Association. “People buy into that which they have a hand in crafting.”
Common Sense’s Buch continued to bring the focus back to students and ensuring their safety, not just through security protocols, but through a clear vision of what students should be learning from the product, and evaluating if that is happening.
“I think one lesson we have learned from the last wave of edtech is that adoption is often moving faster than the fundamental questions we should be asking ourselves. What problems are we solving, and does this actually improve learning? Is this right for the kids at this particular age? And with AI, we have a chance to ask those questions up front and now,” Buch said.
“So, when I think about bans and moratoriums — and that it is such a hot topic — I think there is something to be said about putting a pause until we understand and can move forward in a way that is evidence-based, or where we have a correct plan.”
Resources
CSBA’S AI Taskforce and resources
The AI Taskforce: Education in the Age of Artificial Intelligence was created to help guide school decision-makers in the new technology. Resources include resources and news, scenarios and resolutions and promising practices and policies, along with webinars and relevant publications. If the CSBA website is down, please see this blog post.
National AI Safety & Privacy Standard for Schools
This document can be used as a framework and memorandum of agreement when evaluating and purchasing edtech products
K-12 Generative AI Readiness Checklist Questionnaire
This resource from the CGCS and the Consortium for School Networking (CoSN), in partnership with Amazon Web Services, was developed to help school districts evaluate their preparedness before deploying generative artificial intelligence in classrooms and operations.
CoSN/CGCS K-12 Gen AI Maturity Tool
This tool empowers school districts to assess their preparedness for the responsible integration of Generative AI technologies to support their instructional and operational objectives. It provides tailored recommendations for enhancement across key domain areas, facilitating ongoing improvement efforts.
AI Toolkit for School Districts
This toolkit from Common Sense Media was designed to help ensure that AI use is aligned with a district’s unique mission, vision and goals.
