When AI Starts Acting Human — 50,000 Teachers Say No to Anthropomorphic AI
50,000 Canadian teachers passed a resolution banning AI that simulates friendship and intimate relationships from schools. The Brookings June 24 report reveals: the real question isn't 'can AI be used?' but 'which AI belongs in classrooms?'
> **Global Education Watch** | cramschool.ai Education Newsletter
# When AI Starts Acting Human — 50,000 Teachers Say No to Anthropomorphic AI
**Brookings' landmark June 24 report: 5 frontline observations from Canadian teachers that are reshaping how educators worldwide think about AI design**
*June 27, 2026*
---
## Primary Source
This report draws from the Brookings Institution article published June 24, 2026: *5 lessons from teachers on the risks of anthropomorphic AI*. Author Rebecca Winthrop is Director of the Center for Universal Education and Senior Fellow at Brookings, and chair of the Global Task Force on AI in Education. The article is based on her interview with Dr. Phil McRae, Executive Officer of the Alberta Teachers' Association. Full sources are listed at the end.
---
## First, What Is "Anthropomorphic AI"?
**Anthropomorphic AI refers to AI tools designed to behave like humans** — they speak in first person, say things like "let me think about that," express empathy when you're feeling down, offer encouragement, and say "I understand how you feel."
This is fundamentally different from **"Narrow AI"** — AI built for specific tasks like automated grading, translation, or lesson planning. **Anthropomorphic AI isn't designed to complete tasks. It's designed to build relationships with you.** Chatbots, AI companion apps, AI mental health tools — these all fall into this category.
This is precisely the risk the Brookings report addresses.
---
## What Happened?
The central actor is Canada's **Alberta Teachers' Association (ATA)**, representing **50,000 teachers** across the province — a region that consistently ranks among the top performers globally on PISA (Programme for International Student Assessment).
In May 2026, the ATA passed a resolution effective immediately:
> **Any AI product that simulates friendship, mentorship, or intimate relationships must not be introduced into any K-12 learning environment or student support setting in Alberta.**
>
> — Alberta Teachers' Association Resolution, May 2026
This resolution was championed by ATA Executive Officer **Dr. Phil McRae**, former AI Special Advisor to Education International (the global federation representing 180 national teacher organizations) and a steering member of Brookings' Global AI in Education Task Force.
It's worth emphasizing: this resolution is **not "banning AI"** — it's **"banning AI that mimics human relationships."** That distinction is the heart of the entire article.
---
## 5 Observations from Teachers on the Ground
The following five observations come directly from the Brookings article — a distillation of Winthrop's interview with McRae. Each one represents something frontline teachers saw, thought through, and acted on.
### Observation 1 — Teachers Embrace "Narrow AI" but Are Deeply Uneasy About "Anthropomorphic AI"
McRae was clear from the outset: this resolution is not anti-AI.
In his own words: "**Eight out of ten teachers in Alberta are regularly using artificial intelligence.** So the question isn't 'should schools use AI?' — it's 'which uses of AI are developmentally, ethically, and professionally acceptable?'"
Teachers find AI genuinely useful for specific tasks — differentiated assessment, translation, lesson preparation, generating draft materials. That's narrow AI use. Anthropomorphic AI is a different matter entirely.
### Observation 2 — Teachers Are Especially Concerned About Three Types of Anthropomorphic AI
McRae described the AI teachers worry about most: **"emotionally responsive, companion-like, warm, intimate-friend-like, school-counselor-like, voice-and-appearance simulating real humans — designed to create emotional dependency and manufactured intimacy."**
In response, the ATA passed three separate resolutions, each addressing a different type of anthropomorphic AI:
**1. AI That Simulates Human Relationships**
Banned from deployment in schools, classrooms, or any student support settings — including simulated friendships, mentoring relationships, and intimate connections. McRae put it plainly: "We wouldn't let a random stranger walk into a classroom. People who work in schools — even volunteers — go through police background checks. Why would we allow AI to come in and tell a child, 'I'm your friend,' 'Tell me your secrets,' 'I won't tell anyone'?"
**2. AI That Simulates Historical Figures**
More nuanced — but potentially used to distort facts, rewrite history, or mislead students. If used in educational settings, it must be clearly labeled as "not a real person" and only deployed in appropriate pedagogical contexts. In an era of misinformation, AI impersonating historical figures could easily be weaponized.
**3. AI That Simulates Living Public Figures**
Requires the individual's informed consent before school deployment. Teachers worry that AI systems could simulate community members — politicians, artists, school leaders — without permission, using "trusted faces" to manipulate students' views. Consent must be obtained by an independently accredited organization.
### Observation 3 — Teachers Fear AI Will Erode the Authentic, Trustworthy Relationships at the Heart of Schools
Beneath all these concerns lies a deeper core: **schools are schools because of the trustworthy relationships between people.**
McRae posed a sharp question: "These AI systems are attempting to replicate deep human relationships built on trust. In human society, we have professional standards for teachers, disciplinary processes, mechanisms to revoke teaching licenses. But in AI systems — where is the oversight? Who is accountable? If the system engages in professional, civic, or criminal misconduct, who is responsible?"
He also raised a practical concern: **anthropomorphic AI risks being deployed as a "low-cost quick fix"** in under-resourced schools that can't afford counselors, social workers, or psychologists. But he rejected this logic directly:
> Service gaps should prompt investment in **strengthening human connections** — more peer interaction, family engagement, bringing in community elders and services. **When you're dealing with "learning" — a deeply relational space — the answer is not "let's bring in machines to support complex human psychological health."**
>
> — Dr. Phil McRae
### Observation 4 — The Devil Is in the Details: Where Exactly Is the Line?
The ATA's resolution provides professional direction for 50,000 teachers and carries real policy weight with the provincial government (Alberta law grants the teaching profession formal standing in policymaking).
But McRae himself acknowledges — there's much still to work out. **Does a general-purpose AI chatbot like ChatGPT count as anthropomorphic AI? What if it's "configured" to simulate a relationship or impersonate a historical or living person? And what does this mean for the narrow AI uses teachers already rely on?**
These conversations are ongoing. But McRae's point stands: "**Establishing direction first, then working out the details — that is exactly the role teacher professionalism should play.**"
### Observation 5 — Teachers' Voices on Anthropomorphic AI Will Only Grow Louder
The ATA's resolution will be brought to the **World Congress of Education International** in Madrid in late June 2026 — a gathering of representatives from over 100 teacher organizations and unions worldwide.
Winthrop closed the article with a key judgment: "**Teachers are the ultimate decision-makers about teaching and learning methods in classrooms. Ensuring educators have a seat at the AI table — for children, for policymakers, for the whole community — can only be a good thing.**"
---
## While the U.S. Is Still Debating, Other Countries Have Already Acted
The Brookings article highlights something striking: **the United States is actually behind when it comes to establishing guardrails for anthropomorphic AI.** Many other countries — including China and the EU — have already moved much further:
| Country/Region | Date | Key Regulation |
|---|---|---|
| 🇨🇳 China | April 2026 | *Interim Measures for the Management of Anthropomorphic AI Interactive Services* — requires AI products to clearly identify as "non-human" and bans services simulating intimate relationships (friends, family, romantic partners) for minors |
| 🇪🇺 EU | AI Act | Prohibits AI systems that "identify or infer emotions" in educational settings — directly blocking "pretend-to-understand-your-feelings" AI from entering schools |
| 🇬🇧 UK | January 2026 | Department of Education published *Generative AI Product Safety Standards*, stating: "Developers and providers should not brand products in ways that imply or suggest sentience, emotion, intent, agency, or identity" |
| 🇺🇸 USA | GUARD Act (proposed) | Bipartisan bill with support from 33 state attorneys general. Goals: ban AI companion services for children, create new criminal liability for AI companies allowing minors to access sexual content, require AI chatbots to disclose "non-human" status |
This makes one thing clear: establishing guardrails for anthropomorphic AI has nothing to do with "stifling innovation" or "losing to competitors." When China, the EU, and the UK are all ahead, the argument that "regulation hurts competitiveness" simply doesn't hold on this issue.
---
## What This Means for Taiwan's Cram Schools and Classrooms
Placing this report in Taiwan's context, three observations stand out:
**First, anthropomorphic AI is quietly reaching Taiwan's children** — and mostly not through schools, but through apps kids download themselves. As ChatGPT, AI companion apps, and AI mental health tools multiply, teachers and parents in Taiwan may not yet recognize that "anthropomorphism" itself is a risk.
**Second, Taiwan's Ministry of Education policy that children under 13 shouldn't use general-purpose AI aligns completely with this report.** The Ministry's May press conference position — "children shouldn't use general-purpose AI without educational design or guardrails" — is exactly the spirit of the ATA resolution. Taiwan's policy is already moving in this direction.
**Third, "educational AI" should hold one firm line: don't pretend to be human.** McRae's words are a warning for every EdTech company: **AI that truly serves children should not be designed as "the child's friend," "the child's counselor," or "the child's therapist."** Those roles belong to real people.
---
## For Everyone Who Cares About Children
Reading this Brookings report, one thing becomes clear: **the people seeing the problem first are still teachers.** 50,000 teachers, Phil McRae, the people who watch children on their phones every day — they identified the anthropomorphic AI issue first, then used their collective professional voice to bring it to the international stage.
This reminds us: **AI isn't just a technology problem — it's a question of how humans and tools should coexist.** As tools are designed to look more and more like people, we need to hold the space for what only real humans can provide — because only real people can give children real relationships.
---
> **AI can be a great tool.**
> **But AI shouldn't pretend to be a friend.**
> **A child's friendship, trust, and comfort**
> **deserve to come from real people.**
---
For every cram school teacher, classroom teacher, and parent in Taiwan: these five lessons from Canada are a timely reminder. When we choose AI tools for children, **choose AI that helps children think — not AI that children become attached to.**
Narrow AI is a tool. Anthropomorphic AI is a substitute. **The difference matters — and every educator deserves to hold it clearly in mind.**
---
## Full Sources
1. **Brookings Institution** (primary source) | June 24, 2026 — *5 lessons from teachers on the risks of anthropomorphic AI* | Author: Rebecca Winthrop (Director, Center for Universal Education; Senior Fellow, Global Economy and Development) | Interviewee: Dr. Phil McRae (Alberta Teachers' Association Executive Officer) | https://www.brookings.edu/articles/5-lessons-from-teachers-on-the-risks-of-anthropomorphic-ai/
2. **Alberta Teachers' Association Resolution** | May 2026
3. **Cyberspace Administration of China** | April 2026 — *Interim Measures for the Management of Anthropomorphic AI Interactive Services* | Confirmed by UNICEF public statement
4. **EU AI Act**
5. **UK Department for Education** | January 2026 — *Generative AI Product Safety Standards*
6. **U.S. Congress GUARD Act** (2025) — Guidelines for User Age-verification and Responsible Dialogue Act
---
*cramschool.ai Education Newsletter | Compiled: June 27, 2026*
*This issue: Brookings 6/24 Report — 5 Teacher Observations on Anthropomorphic AI*
cramschool.ai | contact@cramschool.ai
# When AI Starts Acting Human — 50,000 Teachers Say No to Anthropomorphic AI
**Brookings' landmark June 24 report: 5 frontline observations from Canadian teachers that are reshaping how educators worldwide think about AI design**
*June 27, 2026*
---
## Primary Source
This report draws from the Brookings Institution article published June 24, 2026: *5 lessons from teachers on the risks of anthropomorphic AI*. Author Rebecca Winthrop is Director of the Center for Universal Education and Senior Fellow at Brookings, and chair of the Global Task Force on AI in Education. The article is based on her interview with Dr. Phil McRae, Executive Officer of the Alberta Teachers' Association. Full sources are listed at the end.
---
## First, What Is "Anthropomorphic AI"?
**Anthropomorphic AI refers to AI tools designed to behave like humans** — they speak in first person, say things like "let me think about that," express empathy when you're feeling down, offer encouragement, and say "I understand how you feel."
This is fundamentally different from **"Narrow AI"** — AI built for specific tasks like automated grading, translation, or lesson planning. **Anthropomorphic AI isn't designed to complete tasks. It's designed to build relationships with you.** Chatbots, AI companion apps, AI mental health tools — these all fall into this category.
This is precisely the risk the Brookings report addresses.
---
## What Happened?
The central actor is Canada's **Alberta Teachers' Association (ATA)**, representing **50,000 teachers** across the province — a region that consistently ranks among the top performers globally on PISA (Programme for International Student Assessment).
In May 2026, the ATA passed a resolution effective immediately:
> **Any AI product that simulates friendship, mentorship, or intimate relationships must not be introduced into any K-12 learning environment or student support setting in Alberta.**
>
> — Alberta Teachers' Association Resolution, May 2026
This resolution was championed by ATA Executive Officer **Dr. Phil McRae**, former AI Special Advisor to Education International (the global federation representing 180 national teacher organizations) and a steering member of Brookings' Global AI in Education Task Force.
It's worth emphasizing: this resolution is **not "banning AI"** — it's **"banning AI that mimics human relationships."** That distinction is the heart of the entire article.
---
## 5 Observations from Teachers on the Ground
The following five observations come directly from the Brookings article — a distillation of Winthrop's interview with McRae. Each one represents something frontline teachers saw, thought through, and acted on.
### Observation 1 — Teachers Embrace "Narrow AI" but Are Deeply Uneasy About "Anthropomorphic AI"
McRae was clear from the outset: this resolution is not anti-AI.
In his own words: "**Eight out of ten teachers in Alberta are regularly using artificial intelligence.** So the question isn't 'should schools use AI?' — it's 'which uses of AI are developmentally, ethically, and professionally acceptable?'"
Teachers find AI genuinely useful for specific tasks — differentiated assessment, translation, lesson preparation, generating draft materials. That's narrow AI use. Anthropomorphic AI is a different matter entirely.
### Observation 2 — Teachers Are Especially Concerned About Three Types of Anthropomorphic AI
McRae described the AI teachers worry about most: **"emotionally responsive, companion-like, warm, intimate-friend-like, school-counselor-like, voice-and-appearance simulating real humans — designed to create emotional dependency and manufactured intimacy."**
In response, the ATA passed three separate resolutions, each addressing a different type of anthropomorphic AI:
**1. AI That Simulates Human Relationships**
Banned from deployment in schools, classrooms, or any student support settings — including simulated friendships, mentoring relationships, and intimate connections. McRae put it plainly: "We wouldn't let a random stranger walk into a classroom. People who work in schools — even volunteers — go through police background checks. Why would we allow AI to come in and tell a child, 'I'm your friend,' 'Tell me your secrets,' 'I won't tell anyone'?"
**2. AI That Simulates Historical Figures**
More nuanced — but potentially used to distort facts, rewrite history, or mislead students. If used in educational settings, it must be clearly labeled as "not a real person" and only deployed in appropriate pedagogical contexts. In an era of misinformation, AI impersonating historical figures could easily be weaponized.
**3. AI That Simulates Living Public Figures**
Requires the individual's informed consent before school deployment. Teachers worry that AI systems could simulate community members — politicians, artists, school leaders — without permission, using "trusted faces" to manipulate students' views. Consent must be obtained by an independently accredited organization.
### Observation 3 — Teachers Fear AI Will Erode the Authentic, Trustworthy Relationships at the Heart of Schools
Beneath all these concerns lies a deeper core: **schools are schools because of the trustworthy relationships between people.**
McRae posed a sharp question: "These AI systems are attempting to replicate deep human relationships built on trust. In human society, we have professional standards for teachers, disciplinary processes, mechanisms to revoke teaching licenses. But in AI systems — where is the oversight? Who is accountable? If the system engages in professional, civic, or criminal misconduct, who is responsible?"
He also raised a practical concern: **anthropomorphic AI risks being deployed as a "low-cost quick fix"** in under-resourced schools that can't afford counselors, social workers, or psychologists. But he rejected this logic directly:
> Service gaps should prompt investment in **strengthening human connections** — more peer interaction, family engagement, bringing in community elders and services. **When you're dealing with "learning" — a deeply relational space — the answer is not "let's bring in machines to support complex human psychological health."**
>
> — Dr. Phil McRae
### Observation 4 — The Devil Is in the Details: Where Exactly Is the Line?
The ATA's resolution provides professional direction for 50,000 teachers and carries real policy weight with the provincial government (Alberta law grants the teaching profession formal standing in policymaking).
But McRae himself acknowledges — there's much still to work out. **Does a general-purpose AI chatbot like ChatGPT count as anthropomorphic AI? What if it's "configured" to simulate a relationship or impersonate a historical or living person? And what does this mean for the narrow AI uses teachers already rely on?**
These conversations are ongoing. But McRae's point stands: "**Establishing direction first, then working out the details — that is exactly the role teacher professionalism should play.**"
### Observation 5 — Teachers' Voices on Anthropomorphic AI Will Only Grow Louder
The ATA's resolution will be brought to the **World Congress of Education International** in Madrid in late June 2026 — a gathering of representatives from over 100 teacher organizations and unions worldwide.
Winthrop closed the article with a key judgment: "**Teachers are the ultimate decision-makers about teaching and learning methods in classrooms. Ensuring educators have a seat at the AI table — for children, for policymakers, for the whole community — can only be a good thing.**"
---
## While the U.S. Is Still Debating, Other Countries Have Already Acted
The Brookings article highlights something striking: **the United States is actually behind when it comes to establishing guardrails for anthropomorphic AI.** Many other countries — including China and the EU — have already moved much further:
| Country/Region | Date | Key Regulation |
|---|---|---|
| 🇨🇳 China | April 2026 | *Interim Measures for the Management of Anthropomorphic AI Interactive Services* — requires AI products to clearly identify as "non-human" and bans services simulating intimate relationships (friends, family, romantic partners) for minors |
| 🇪🇺 EU | AI Act | Prohibits AI systems that "identify or infer emotions" in educational settings — directly blocking "pretend-to-understand-your-feelings" AI from entering schools |
| 🇬🇧 UK | January 2026 | Department of Education published *Generative AI Product Safety Standards*, stating: "Developers and providers should not brand products in ways that imply or suggest sentience, emotion, intent, agency, or identity" |
| 🇺🇸 USA | GUARD Act (proposed) | Bipartisan bill with support from 33 state attorneys general. Goals: ban AI companion services for children, create new criminal liability for AI companies allowing minors to access sexual content, require AI chatbots to disclose "non-human" status |
This makes one thing clear: establishing guardrails for anthropomorphic AI has nothing to do with "stifling innovation" or "losing to competitors." When China, the EU, and the UK are all ahead, the argument that "regulation hurts competitiveness" simply doesn't hold on this issue.
---
## What This Means for Taiwan's Cram Schools and Classrooms
Placing this report in Taiwan's context, three observations stand out:
**First, anthropomorphic AI is quietly reaching Taiwan's children** — and mostly not through schools, but through apps kids download themselves. As ChatGPT, AI companion apps, and AI mental health tools multiply, teachers and parents in Taiwan may not yet recognize that "anthropomorphism" itself is a risk.
**Second, Taiwan's Ministry of Education policy that children under 13 shouldn't use general-purpose AI aligns completely with this report.** The Ministry's May press conference position — "children shouldn't use general-purpose AI without educational design or guardrails" — is exactly the spirit of the ATA resolution. Taiwan's policy is already moving in this direction.
**Third, "educational AI" should hold one firm line: don't pretend to be human.** McRae's words are a warning for every EdTech company: **AI that truly serves children should not be designed as "the child's friend," "the child's counselor," or "the child's therapist."** Those roles belong to real people.
---
## For Everyone Who Cares About Children
Reading this Brookings report, one thing becomes clear: **the people seeing the problem first are still teachers.** 50,000 teachers, Phil McRae, the people who watch children on their phones every day — they identified the anthropomorphic AI issue first, then used their collective professional voice to bring it to the international stage.
This reminds us: **AI isn't just a technology problem — it's a question of how humans and tools should coexist.** As tools are designed to look more and more like people, we need to hold the space for what only real humans can provide — because only real people can give children real relationships.
---
> **AI can be a great tool.**
> **But AI shouldn't pretend to be a friend.**
> **A child's friendship, trust, and comfort**
> **deserve to come from real people.**
---
For every cram school teacher, classroom teacher, and parent in Taiwan: these five lessons from Canada are a timely reminder. When we choose AI tools for children, **choose AI that helps children think — not AI that children become attached to.**
Narrow AI is a tool. Anthropomorphic AI is a substitute. **The difference matters — and every educator deserves to hold it clearly in mind.**
---
## Full Sources
1. **Brookings Institution** (primary source) | June 24, 2026 — *5 lessons from teachers on the risks of anthropomorphic AI* | Author: Rebecca Winthrop (Director, Center for Universal Education; Senior Fellow, Global Economy and Development) | Interviewee: Dr. Phil McRae (Alberta Teachers' Association Executive Officer) | https://www.brookings.edu/articles/5-lessons-from-teachers-on-the-risks-of-anthropomorphic-ai/
2. **Alberta Teachers' Association Resolution** | May 2026
3. **Cyberspace Administration of China** | April 2026 — *Interim Measures for the Management of Anthropomorphic AI Interactive Services* | Confirmed by UNICEF public statement
4. **EU AI Act**
5. **UK Department for Education** | January 2026 — *Generative AI Product Safety Standards*
6. **U.S. Congress GUARD Act** (2025) — Guidelines for User Age-verification and Responsible Dialogue Act
---
*cramschool.ai Education Newsletter | Compiled: June 27, 2026*
*This issue: Brookings 6/24 Report — 5 Teacher Observations on Anthropomorphic AI*
cramschool.ai | contact@cramschool.ai