When AI Defaults to Using Your Face: A Mandatory Lesson on Consent

Last week in IT class, a student held up their phone and asked me, "Mr. Shen, my classmate used AI to turn my photo into a meme and posted it in the group chat. Can I sue them?" Before the words fell, another student interjected, "But Meta's new feature says that as long as photos are public on Instagram, others can use AI to generate your likeness." The classroom instantly erupted.

I realized this was exactly the AI ethics teaching case I had been looking for. In early July, Meta launched a new AI feature allowing users to tag public Instagram accounts via the AI App and generate images using their portraits. Although Meta claims it checks for policy violations, the core mechanism is: default consent unless you actively opt out.

This feature perfectly illustrates a concept I repeatedly emphasize to students: the huge gap between "default" and "choice." As an IT teacher, I must break this down into layers students can understand.

Question 1: What Filter Did Technology Add to "Consent"?

Traditionally, using someone else's portrait required explicit authorization—signatures, confirmations, mutual awareness. But in the AI era, this logic is being rewritten. Meta's approach is: as long as the account is public, the AI model can use it as training material and allow other users to generate derivative content based on that portrait. The user's only recourse is to manually "opt out" in settings.

Let's look at the numbers:

Dimension	Traditional Portrait Use	Meta AI New Feature
Default State	Requires active authorization	Automatic authorization (opt-out)
Awareness Level	Users explicitly know the purpose	Users may be completely unaware
Revocation Cost	Must go through platform review	Manually find the settings switch
Scope of Application	Single use	Unlimited AI generations

I had students conduct an experiment: Find the entry point for Meta's privacy settings. Result: Only 3 out of 35 could complete it within 5 minutes. And that was just finding the switch, let alone users who didn't notice the new feature notification at all. When "informed consent" becomes "struggling to object after being informed," it is effectively stripping control from ordinary users.

Question 2: What Does This Mean for AI Education?

Many tech companies promote AI by emphasizing "enabling everyone to create," ignoring that "enabling everyone to refuse" is equally important. In one of my innovation projects, a group of students wanted to use an open-source AI model to generate avatars of classmates for a virtual exhibition. My first reaction was: "Did you get consent forms?" They froze.

This case perfectly introduces three teaching points:

• Data Provenance Education: Where does the training data for AI models come from? Is it actively contributed by users or passively scraped? Students need to understand the legality of training data sources.
• Asymmetry of Technology: Big companies have legal teams and technical resources to optimize user experience; ordinary users do not. Rights with steeper learning curves are more easily ignored.
• Mock Court Exercise: I divided students into groups playing users, AI companies, and regulators, discussing "Does public data equal commercially usable data?" Results showed that most students initially believed "public = free to use," but changed their views after discussion.

This isn't just moral preaching. From a technical perspective, Meta's feature relies on image embedding models and diffusion models. I often tell students: "If you can use code to generate someone else's face, you must be able to use code to protect someone else's face." In class, I guide students to compare how different AI platforms explain user data usage, creating a "Consent Cost Ranking Table":

Platform	Notification Method	Exit Difficulty	Teaching Score (Lower is Worse)
Meta AI	Feature update announcement	Multi-layer menus	2/10
OpenAI	Account settings + email notification	Single checkbox	5/10
An Open Source Model	Public license agreement	User self-check	1/10

This comparison intuitively shows students: Not all "consents" are created equal.

Question 3: What Can We Teach Students?

Since we cannot stop big companies from launching such features, as a teacher, what I can give students is not panic, but tools. I designed a small project assignment—"AI Portrait Protection Guide." Students need to complete the following steps:

1. Find the "Data Usage Settings" entry for the 3 AI platforms they use most, and screenshot the path.

2. Describe in a flowchart "What rights do I have when my portrait is AI-generated, and how do I exercise them."

3. Write an explanation of no more than 200 words for parents explaining "Why proactively opting out" is better than trusting default settings.

After completing the assignment, students universally reported the same keyword: Exhaustion. One student wrote in their summary: "Mr. Shen, just checking the exit option for Meta took 20 minutes, clicking through layers, and filling in reasons for two fields. This design itself is meant to deter people." This is the core of the problem—when the "opt-out" process is complex enough to make users give up resisting, the so-called "choice" exists in name only.

I am not against AI technology; on the contrary, I believe it stimulates creativity. But as an educator, I must make students understand: Technology's default settings are not morally neutral zones, but declarations of values. Meta's feature may be commercially clever, but in the educational dimension, it provides us with the most realistic case study: How much effort does a person need to expend to reclaim the right to "not be used" when AI can casually use your face?

One-sentence summary: The core literacy of the AI era is not knowing how to use AI to generate images, but knowing when and how to refuse being generated by AI.

Original Link: https://www.wired.com/story/please-stop-making-me-opt-out-of-ai/