Regulations Hitting AI Usage in EdTech

Was this forwarded to you by a friend? Sign up, and get your own copy of the news that matters sent to your inbox every week. Sign up for the On EdTech newsletter. Interested in additional analysis? Try with our 30-day free trial and Upgrade to the On EdTech+ newsletter.

Late last week Van Davis wrote at the WCET blog about how AI usage at colleges and universities should be designed with regulations in mind, particularly Regular and Substantive Interaction (RSI).

After describing RSI regulatory language, Davis noted that the final 2020 regulations on RSI specifically called out technology usage such as AI [emphasis added].

Although the milestone release of ChatGPT as a commercial product was just 14 months ago, that does not mean that government regulatory efforts are waiting until we have clarity on real use cases or impacts. While the AI genie is out of the bottle, there are major efforts to fence it in, often in naive fashion (it’s hard to build a fence around something which is not understood).

US Executive Order

In late October 2023, President Biden issued an 80-page executive order on AI that does not constitute regulatory action, per se, but that directs government agencies to create regulations and guidance. The headlines from the fall centered on the item listed in the executive order’s fact sheet.

Resources to help educators, sounds hard to argue with that. That aim, however, is but one part of the broader regulatory agenda sought in the actual executive order.

There are several sections based on helping educators, and reminding people that consumer protections will apply to AI-impacted industries, but for EdTech the biggest section is the following (section 8 d) [emphasis added].

In other words, the executive order is directing the Secretary of Education to produce regulatory guidance by next fall - not actual regulations going through negotiated rulemaking, or heaven forbid making statutory changes - with nondiscrimination as the core element. Basically, be prepared, something is coming down the road.

The second piece is interesting as it essentially calls for implementing recommendations from this May 2023 report. That 71-page report was based on “constituent meetings” held in June and August, 2022. aka several months before the true (or at least visible) dawning of generative AI. The report’s intro - which again will form the basis of much of the upcoming guidance - has a strangely outdated and limited definition of the new AI capabilities [emphasis added].

Apparently AI-enabled systems change so quickly that in May 2023 ED was not aware that generative AI goes beyond chatbots.

The seven recommendations from the report:

1. Emphasize Humans in the Loop

2. Align AI Models to a Shared Vision for Education

3. Design Using Modern Learning Principles

4. Prioritize Strengthening Trust

5. Inform and Involve Educators

6. Focus R&D on Addressing Context and Enhancing Trust and Safety

7. Develop Education-Specific Guidelines and Guardrails

Each recommendation includes a 3 - 10 paragraph section with general discussions of the very general points. In fact, it is hard to really call these recommendations rather than semi-organized aspirational sentiments that are hard to argue with. But viewed another way is that the chain from the May 2023 report to the October 2023 executive order to the future 2024 guidance represents effectively a blank check to craft Dear Colleague Letters and saying hey, I was told by the President to do this.

The one recommendation with a little bit of meat on the bones is the Shared Vision from #2. This gives a conceptual view of AI-based tools.

EU: Hold my Belgian Tripel

While the Biden Executive Order and referenced AI report constitute a lot of pages pushing regulatory guidance on tools without significant use cases yet, the European Union (EU) and the UK have gone further.

The EU created the AI Act in August 2023, and in December the Parliament and Council came to agreement on the terms. The AI Act is now headed towards formal commitee votes and formal adoption before becoming law.

Unlike the US executive order, the EU’s AI Act has more specifics based on a risk pyramid, as shown below.

There are specific banned usage, including social scoring and harmful manipulative 'subliminal techniques', and there are different levels of risk.

What is important to understand for the vendor and investment community is the scope of regulations. As described in Telefónica:

GDPR, DMA, and DSA represent a family of regulatory actions that each had surprising (and confusing scope) and a host of unintended consequences. The AI Act will likely have strong similarities.

Expect Actions in 2024

I don’t pretend to be able to describe all regulatory activities around AI that could impact EdTech. The bigger issue is for readers to be aware that 2024 is shaping up to be a banner year in the introduction of new laws and regulations and regulatory guidance around AI, and that these actions will impact education. And as the WCET post has shown, there are already regulatory limitations that the EdTech community should understand.

Stay tuned.

The main On EdTech newsletter is free to share in part or in whole. All we ask is attribution.

Thanks for being a subscriber.