The emergency in education now
The AI education emergency
Our kids need us to move forward. Together. Now.
Countries are rebuilding school around artificial intelligence right now. China, the UAE and India have ordered AI into their school curricula. In the United States, a record share of 8th graders read below the most basic level the national test measures. AI can teach a lesson. Only people can teach a person. This is the one issue we move forward together, now.
- AI teaches the lesson. Teachers teach the person.
- A first grader this fall graduates in 2038.
- Elections come every two years. A school year never comes back.
The world is moving
Governments on four continents have put AI into school by law, plan or curriculum. Each entry is the government's own action, dated, with its source.
China
China's Ministry of Education General Office issued the Notice on Strengthening AI Education in Primary and Secondary Schools (Jiao Ji Ting Han [2024] No. 32).
Shenzhen Municipal Education Bureau (government explainer of the MOE notice) szeb.sz.gov.cn
More, and the source text
Details: Six main tasks: a systematic curriculum, regular teaching and evaluation, general teaching resources, ubiquitous teaching environments, teacher supply at scale, and exchange activities. Grade bands, per the MOE text carried by Xinhua on 2024-12-02 (https://www.xinhuanet.com/edu/20241202/4975d14d75294fe8989632232a9a2c89/c.html): 小学低年级段侧重感知和体验人工智能技术，小学高年级段和初中阶段侧重理解和应用人工智能技术，高中阶段侧重项目创作和前沿应用 (lower primary focuses on perceiving and experiencing AI; upper primary and junior high on understanding and applying it; senior high on project creation and frontier applications). China Education Daily (2024-12-18, reprinted at https://www.tsinghua.edu.cn/info/1182/116034.htm) reports the notice sets a deadline: 2030年前在中小学基本普及人工智能教育 (basically universalize AI education in primary and secondary schools before 2030).
The source says: "教基厅函〔2024〕32号 ... 2024年11月20日 (Translation: Document Jiao Ji Ting Han [2024] No. 32 ... November 20, 2024)"
China (Beijing)
The Beijing Municipal Education Commission ordered every primary and secondary school in the city to teach general AI education from the fall 2025 semester, at least 8 class hours per school year.
Beijing Municipal Education Commission jw.beijing.gov.cn
More, and the source text
Details: Set in the Beijing Work Plan for Promoting AI Education in Primary and Secondary Schools (2025 to 2027), written to carry out MOE notice No. 32 of 2024. Courses may be taught standalone or integrated with information technology and other subjects (Xinhua, via english.www.gov.cn, 2025-03-12). Plan made public 2025-03-07. On 2025-06-26 the commission followed with the Beijing Local Curriculum Outline for AI Education in Primary and Secondary Schools (Trial, 2025 Edition): https://jw.beijing.gov.cn/xxgk/2024zcwj/2024qtwj/202506/t20250626_4127392.html
The source says: "从2025年秋季学期开始，全市中小学校开展人工智能通识教育，每学年不少于8课时，实现中小学生全面普及。 (Translation: Starting in the fall 2025 semester, all primary and secondary schools in the city will provide general AI education, no fewer than 8 class hours per school year, reaching every primary and secondary student.)"
China
Five Chinese agencies led by the Ministry of Education issued the AI+Education Action Plan, ordering AI education into every local curriculum with set goals, content and class hours for each school stage.
Government of China (gov.cn) gov.cn
More, and the source text
Details: Document 教科信〔2026〕1号, issued by the Ministry of Education, National Development and Reform Commission, Ministry of Industry and Information Technology, Ministry of Science and Technology and National Data Administration. Orders schools to 开齐开足开好人工智能相关课程 (offer AI-related courses in full). Target: 到2030年，人工智能与教育深度融合格局基本形成 (by 2030, deep integration of AI and education is basically in place). Science and Technology Daily (2026-04-10, https://www.stdaily.com/web/gdxw/2026-04/10/content_500122.html) also quotes: 建强中小学人工智能教育基地，支持农村、边远地区学校利用国家平台开好人工智能课程 (build strong primary and secondary AI education bases; help rural and remote schools teach AI courses through the national platform). No specific hour count is set nationally.
The source says: "推动人工智能教育全面纳入地方课程体系，指导各地研制人工智能课程指南，明确各学段课程目标、内容与课时要求。 (Translation: Bring AI education fully into local curriculum systems, guide each locality to draft AI curriculum guides, and set course goals, content and class-hour requirements for each school stage.)"
China
China's State Council issued the 15th Five-Year Plan for Education Development, ordering AI education across all school stages.
State Council of China (gov.cn) gov.cn
More, and the source text
Details: Document 国发〔2026〕19号, dated 2026-06-22 and published 2026-06-29. It also orders an AI+Education action: 开展“人工智能+教育”行动，推动教育理念、培养模式、考试考核、科研范式和治理机制深层次变革 (carry out the AI+Education action to drive deep change in teaching philosophy, training models, examinations and assessment, research paradigms and governance). The gov.cn text is marked abridged (本文有删减).
The source says: "推进人工智能全学段教育，提升学生人工智能素养，增强提出问题、解决问题的能力。 (Translation: Advance AI education across all school stages, raise students' AI literacy, and strengthen their ability to pose and solve problems.)"
United Arab Emirates
The UAE government approved a final kindergarten to grade 12 curriculum making artificial intelligence a subject in all government schools from the next academic year (2025-2026).
Later: On 2026-09-02 the UAE Cabinet approved the AI curriculum for all public and private schools, with a target of 22,000 teachers trained (see ae-2026-09-02-all-schools).
The National thenationalnews.com
More, and the source text
Details: Announced by Sheikh Mohammed bin Rashid Al Maktoum, who wrote that 'the UAE government today approved the final curriculum' and 'Our goal is to teach our children a deep understanding of AI from a technical perspective.' At the time there was no announcement on private schools, which are regulated separately.
The source says: "artificial intelligence is to be introduced as a subject across all stages of government education in the UAE"
United Arab Emirates
The UAE Cabinet approved implementing the AI curriculum across all public and private schools and set a goal of training 22,000 teachers and educators to use AI in teaching.
Khaleej Times khaleejtimes.com
More, and the source text
Details: Decision taken at a Cabinet meeting chaired by Sheikh Mohammed bin Rashid Al Maktoum. The article does not state the academic year in which private schools begin.
The source says: "The UAE has approved the implementation of an artificial intelligence curriculum across all public and private schools"
South Korea
South Korea began classroom use of government-certified AI digital textbooks for elementary grades 3 and 4 and the first year of middle and high school.
Later: On 2025-08-04 the National Assembly amended the Elementary and Secondary Education Act so AI digital textbooks are legally educational materials, not textbooks (see kr-2025-08-04-textbook-status-law). Adoption fell from 37% to 19% in the semester that began in September 2025 (Rest of World, republished by Scroll.in on 2025-11-02: https://scroll.in/article/1087825/scroll_in). On 2025-12-17 the Board of Audit and Inspection reported an average utilization rate of 8.1 percent, measured as use on more than 10 days from March to May 2025 at schools that chose the books, and that 85.5 percent of 513 surveyed English and math teachers at pilot schools had never used them or had stopped (Asia Economy: https://www.asiae.co.kr/en/article/2025121614431525429).
The Korea Times koreatimes.co.kr
More, and the source text
Details: Schools chose whether to adopt them; 3,870 of 11,932 schools (about 32 percent) adopted them as official learning materials (Korea Times, 2025-06-10). The program was a flagship of former President Yoon Suk Yeol; President Lee pledged to redefine the books as supplementary educational materials.
The source says: "Since March, AI-powered digital textbooks have been introduced for third and fourth graders in elementary school,"
South Korea
South Korea's National Assembly passed an amendment to the Elementary and Secondary Education Act that strips AI digital textbooks of textbook status and classifies them as educational materials.
Kyunghyang Shinmun (khan.co.kr) khan.co.kr
More, and the source text
Details: The amendment limits 'textbook' to printed books and e-books and takes effect on promulgation, so AI textbooks can be used only as educational material from the second semester of 2025 (Asia Economy, 2025-08-04: https://www.asiae.co.kr/en/article/2025080418420589912). The Korea Times reported AI materials become 'optional supplementary tools rather than mandatory textbooks' and that an earlier version passed under the Yoon administration was scrapped after a presidential veto.
The source says: "4일 국회 본회의에서 AI 교과서의 지위를 교육자료로 하는 초중등교육법 일부개정안이 통과되면서 AI 교과서는 법률에서 교육자료로 규정된다. (Translation: With the National Assembly plenary session on the 4th passing the partial amendment to the Elementary and Secondary Education Act that makes AI textbooks educational materials, AI textbooks are defined in law as educational materials.)"
Singapore
Singapore's Ministry of Education committed that every student will have foundational AI literacy before entering higher education, under a 'Four Learns' framework.
Ministry of Education, Singapore moe.gov.sg
More, and the source text
Details: Four Learns: 'learn about AI, learn to use AI, learn with AI, and most importantly, learn beyond AI.' Younger students get lower exposure and closer supervision. Stated by Minister of State for Education Jasmin Lau in the Committee of Supply debate.
The source says: "By the time our students enter Institutes of Higher Learning, every one of them will have foundational AI literacy."
Singapore
Singapore's Ministry of Education set tiered rules for AI in schools: no AI-based assignments in Primary 1 to 3, education-specific AI tools under teacher supervision in Primary 4 to 6, and AI tasks allowed in some secondary schools.
Ministry of Education, Singapore moe.gov.sg
More, and the source text
Details: From MOE's combined written answer to parliamentary questions on AI usage in schools. For Primary 1 to 3: 'schools will not assign any work that requires them to use AI directly.'
The source says: "Students in primary 4 to 6 will only use AI tools in school that are specifically designed for education."
India
India's Ministry of Education announced that AI and computational thinking will be taught from Class 3 onward starting in the 2026-27 academic session.
Later: On 2026-04-01 CBSE's Computational Thinking and AI curriculum for Classes 3 to 8 was formally launched (see in-2026-04-01-cbse-ct-ai-launch).
The Tribune (India) tribuneindia.com
More, and the source text
Details: Developed under the National Curriculum Framework for School Education 2023, with the ministry supporting CBSE, NCERT, Kendriya Vidyalaya Sangathan and Navodaya Vidyalaya Sangathan. An expert committee chaired by Prof. Karthik Raman of IIT Madras was formed to write the curriculum. Date is the article's publication date.
The source says: "an AI curriculum will be introduced from Class III onwards beginning next year (the 2026-27 academic session)."
India
India's Education Minister Dharmendra Pradhan launched CBSE's Computational Thinking and AI curriculum for Classes 3 to 8.
All India Radio News (newsonair.gov.in) newsonair.gov.in
More, and the source text
Details: IANS (via Morung Express, 2026-04-01: https://morungexpress.com/cbse-rolls-out-curriculum-on-computational-thinking-ai-for-classes-3-8-from-2026-27) reports it starts in the 2026-27 academic session; Classes 3 to 5 integrate computational thinking across subjects such as mathematics and languages, and Classes 6 to 8 add foundational AI concepts. Pradhan called for the curriculum to be offered in Indian languages.
The source says: "introduce foundational digital and analytical skills to students from Classes III to VIII"
Estonia
Estonia announced AI Leap, a national program to give AI tools and training to high school students and their teachers, planned to launch on 1 September 2025.
Ministry of Education and Research, Estonia hm.ee
More, and the source text
Details: Led by the Ministry of Education and Research with the President's Digital Council and private backers. Planned expansion in September 2026 by 38,000 students and 2,000 teachers as vocational schools and new 10th graders join. Negotiations had begun with OpenAI and Anthropic.
The source says: "20,000 high school students in grades 10-11 and their 3,000 teachers"
Saudi Arabia
Saudi Arabia announced a national AI curriculum for public education from elementary through secondary school, starting in the coming academic year.
Arab News (Pakistan edition), citing the Saudi Press Agency arabnews.pk
More, and the source text
Details: Unveiled by the National Curriculum Center with the Ministry of Education, the Ministry of Communications and Information Technology and the Saudi Data and Artificial Intelligence Authority, per the Saudi Press Agency. Preceded by an 'Introduction to Artificial Intelligence' pilot course for third-year general-track high school students. The article names no year; given its July 2025 date, the coming academic year is 2025-2026.
The source says: "beginning in elementary school and continuing through secondary education"
Japan
Japan's education ministry (MEXT) issued Version 2.0 of its Guidelines on the Use of Generative AI in Primary and Secondary Education.
Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan mext.go.jp
More, and the source text
Details: Ver. 2.0 is a revision of an earlier MEXT guideline. MEXT's AI page lists it as the current primary and secondary school guideline.
The source says: "「初等中等教育段階における生成AIの利活用に関するガイドライン（Ver. 2.0）」（令和6年12月26日改訂） (Translation: Guidelines on the Use of Generative AI at the Primary and Secondary Education Stage (Ver. 2.0), revised December 26, 2024)"
United Kingdom (England)
England's Department for Education published its policy paper on generative AI in education; the current version requires safeguards for any pupil use.
Later: Updated 2023-10-26 (data privacy and intellectual property), 2025-01-22 (intellectual property; link to Generative AI product safety expectations), 2025-06-10 (link to support materials) and 2025-08-12 (Ofsted guidance on AI during inspection), per the GOV.UK change history: https://www.gov.uk/government/publications/generative-artificial-intelligence-in-education
Department for Education (GOV.UK) gov.uk
More, and the source text
Details: The current text (updated 2025-08-12) also states: 'Schools and colleges must not allow or cause students' original work to be used to train generative AI models' and 'We see more immediate benefits and fewer risks from teacher-facing use of generative AI.'
The source says: "Pupils should only be using generative AI in education settings with appropriate safeguards in place"
United Kingdom (England)
The UK government, responding to England's Curriculum and Assessment Review, committed to replace the computer science GCSE with a broader computing GCSE and to explore a new data science and AI qualification for ages 16 to 18.
Department for Education (GOV.UK) gov.uk
More, and the source text
Details: The government will aim to publish the revised national curriculum by spring 2027, for first teaching from September 2028. The data science and AI qualification is at the exploration stage, not a firm commitment.
The source says: "exploring a new qualification in data science and AI for 16–18-year-olds"
European Union
The EU AI Act's prohibitions, including the ban on emotion recognition in education institutions, and its AI literacy obligations began to apply across the EU.
Later: The Commission page states that high-risk obligations for areas including education 'have been extended to 2 December 2027' following the agreement to simplify the AI Act (the Digital Omnibus on AI). The Council adopted that legislative act on 2026-06-29 (Council document 11301/26: https://data.consilium.europa.eu/doc/document/ST-11301-2026-INIT/en/pdf).
European Commission ec.europa.eu
More, and the source text
Details: The European Commission lists 'emotion recognition in workplaces and education institutions' among prohibited practices, and 'AI solutions used in education institutions, that may determine the access to education' as high-risk uses subject to strict obligations.
The source says: "prohibited AI practices and AI literacy obligations entered into application from 2 February 2025"
United States
President Trump signed Executive Order 14277, 'Advancing Artificial Intelligence Education for American Youth,' creating a White House Task Force on AI Education and a Presidential AI Challenge.
The White House whitehouse.gov
More, and the source text
Details: The Task Force is chaired by the Director of the Office of Science and Technology Policy. Orders: Challenge plans within 90 days and the Challenge held within 12 months of the plan; K-12 AI literacy resources built with industry, academic and nonprofit partners, ready within 180 days of the first partnerships; Education Secretary guidance within 90 days on using formula and discretionary grant funds for AI; within 120 days, Education to prioritize AI in teacher-training grants, NSF to prioritize AI-in-education research, USDA to prioritize AI education through 4-H, and Labor to expand AI-related Registered Apprenticeships. Published at 90 FR 17519 on 2025-04-28 (https://www.govinfo.gov/content/pkg/FR-2025-04-28/html/2025-07368.htm).
The source says: "It is the policy of the United States to promote AI literacy and proficiency among Americans"
United States
The White House launched the Pledge to America's Youth: Investing in AI Education, with more than 60 organizations committing four years of resources for youth and teachers.
Later: By 2025-09-04 the White House cited 'the more than 135+ pledges made to support AI education across America' (see us-2025-09-04-task-force-meeting).
The White House whitehouse.gov
More, and the source text
Details: Signers pledge to make available funding, grants, educational materials, technology, teacher training and workforce development, working with the White House Task Force on AI Education.
The source says: "over 60 organizations are the first signers"
United States
The U.S. Department of Education issued a Dear Colleague Letter stating that federal formula and discretionary grant funds may be used for AI uses in education.
U.S. Department of Education (press release mirrored by Public, pubt.io) pubt.io
More, and the source text
Details: The release says the guidance 'outlines how AI may be used across key educational functions' and addresses 'the use of formula and discretionary grant funds to support improved outcomes for learners.' The ed.gov original returned an access error to our fetch; this text is the department's release as distributed by Public (pubt.io).
The source says: "affirms that such uses are allowable under existing federal education programs"
United States
First Lady Melania Trump launched the nationwide Presidential AI Challenge for K-12 students and educators.
The White House whitehouse.gov
More, and the source text
Details: The Challenge was ordered by Executive Order 14277.
The source says: "inviting K-12 students and educators across America to participate"
United States
The U.S. Department of Education finalized its Secretary's Supplemental Priority on Advancing Artificial Intelligence in Education for use across its discretionary grant programs.
Federal Register (govinfo.gov) govinfo.gov
More, and the source text
Details: Published at 91 FR 18774 (FR Doc. 2026-07087). The priority funds projects that, among other things, expand understanding of AI. It joins the Secretary's other supplemental priorities on evidence-based literacy, educational choice, returning education to the states, meaningful learning opportunities and career pathways.
The source says: "The priority and definitions are effective May 13, 2026."
United States (Ohio)
Ohio law (House Bill 96) required every public school district, community school and STEM school to adopt an AI use policy by July 1, 2026.
Ohio Department of Education and Workforce education.ohio.gov
More, and the source text
Details: The same law required the Ohio Department of Education and Workforce to publish a model AI policy by December 31, 2025, which districts may adopt.
The source says: "By July 1, 2026, all Ohio traditional public school districts, community schools, and STEM schools are required"
United States (states)
Thirty-five states and Puerto Rico had issued official guidance or policy on AI use in K-12 schools, per the AI for Education state tracker.
AI for Education aiforeducation.io
More, and the source text
Details: Tracker map marked 'Updated 8/21/2026.' Counts vary by tracker; TeachAI's list showed 30 state entries when fetched. AI for Education is a nonprofit, not a government source.
The source says: "35 states (and Puerto Rico) now have official guidance or policy on the use of AI in K12 schools."
United States (states)
State legislatures enacted 13 bills on AI in classroom instruction during the 2026 session, out of 87 bills in 28 states.
FutureEd (Georgetown University) future-ed.org
More, and the source text
Details: Examples per FutureEd: Alabama H.B. 329 requires a computer science course that includes AI instruction to graduate from high school; Maryland S.B. 720 (Artificial Intelligence Ready Schools Act) requires state AI guidance, district AI policies and coordinators, and paid train-the-trainer professional development; Oklahoma S.B. 1734 requires every district to adopt a written AI policy before the 2027-28 school year and bars AI as the primary basis for grading, discipline or placement. Date is the tracker's last update.
The source says: "So far this session, 13 bills have been enacted, including:"
United States vs China
Georgetown's Center for Security and Emerging Technology projected that Chinese universities would produce more than 77,000 STEM PhD graduates a year by 2025, versus about 40,000 in the United States.
Center for Security and Emerging Technology (CSET), Georgetown University cset.georgetown.edu
More, and the source text
Details: This is a 2021 projection from enrollment patterns, not measured 2025 output. CSET adds that if international students are excluded from the U.S. count, Chinese STEM PhD graduates would outnumber U.S. ones more than three to one.
The source says: "Based on current enrollment patterns, we project that by 2025 Chinese universities will produce more than 77,000 STEM PhD graduates per year"
United States vs China
Stanford HAI's 2026 AI Index reported that the U.S.-China AI model performance gap has effectively closed.
Stanford Institute for Human-Centered AI (HAI) hai.stanford.edu
More, and the source text
Details: Same report: the U.S. still produces more top-tier AI models and higher-impact patents, while China leads in publication volume, citations, patent output and industrial robot installations. U.S. private AI investment reached $285.9 billion in 2025, more than 23 times China's $12.4 billion. The number of AI researchers and developers moving to the U.S. has dropped 89% since 2017. Release date per SiliconANGLE (2026-04-13).
The source says: "The U.S.-China AI model performance gap has effectively closed."
Where our kids stand
The national and international tests, with the year each was given.
40%
About 4 in 10 American 4th graders read below the basic level, the largest share since 2002.
National Assessment Governing Board (NAGB) nagb.gov
More, and the source text
What it measures: Share of U.S. 4th graders below NAEP Basic in reading, 2024 NAEP. Same release: nearly 40% of 8th graders are below NAEP Basic in math, and no state surpassed its 2019 8th grade scores in either subject.
The source says: "Around 40 percent of 4th graders are working below the NAEP Basic level in reading, the largest percentage since 2002."
33%
A third of American 8th graders cannot read at the most basic level the national test measures, the highest share ever recorded.
National Assessment Governing Board (NAGB) nagb.gov
More, and the source text
What it measures: Share of U.S. 8th graders scoring below NAEP Basic in reading, 2024 NAEP (results released January 29, 2025). The page states this is a greater percentage than ever before.
The source says: "A third (33%) of 8th graders are not even reading at the NAEP Basic level"
32%
About 1 in 3 high school seniors read below the basic level, 12 points more than in 1992.
The Nation's Report Card (NCES) nationsreportcard.gov
More, and the source text
What it measures: Share of U.S. 12th graders below NAEP Basic in reading, 2024, up 2 points from 2019 and 12 points from 1992. NAGB's September 9, 2025 release calls it the largest percentage ever below NAEP Basic in 12th grade reading.
The source says: "Thirty-two percent of twelfth-grade students performed in the range of the scale below the NAEP Basic level"
45%
Nearly half of America's high school seniors score below the basic level in math, the highest share ever recorded.
National Assessment Governing Board (NAGB) nagb.gov
More, and the source text
What it measures: Share of U.S. 12th graders below NAEP Basic in mathematics, 2024 NAEP (data collected January to March 2024). NAGB calls it the highest percentage ever to score below that level. Only 22% scored at or above NAEP Proficient.
The source says: "45% of 12th graders scored below NAEP Basic in math"
33%, down from 37% in 2019
Only 1 in 3 high school seniors is ready for entry-level college math.
National Assessment Governing Board (NAGB) nagb.gov
More, and the source text
What it measures: Share of U.S. 12th graders NAGB considers academically prepared for entry-level college math coursework, 2024 versus 2019. In reading, 35% were considered prepared, down from 37%.
The source says: "A third – 33% – of 12th graders were considered academically prepared for entry-level college math coursework in 2024"
14%, down from 35% in 1984
Only 14% of 13-year-olds read for fun almost every day, down from 35% in 1984.
National Assessment Governing Board (NAGB) nagb.gov
More, and the source text
What it measures: Share of U.S. 13-year-olds reporting reading for fun almost every day, NAEP Long-Term Trend 2025 (released June 10, 2026). Same release: 13-year-olds' scores were unchanged from 2023 and still below 2020; 9-year-olds gained 4 points in both reading and math versus 2022.
The source says: "Only 14 percent of 13-year-olds report reading for fun almost every day in 2025"
490, down from 504 in 2022
American 15-year-olds' reading score fell 14 points in three years on the main international test.
National Center for Education Statistics (NCES) nces.ed.gov
More, and the source text
What it measures: U.S. average PISA reading score, 15-year-olds, 2025 (490) versus 2022 (504). Still above the 2025 OECD average of 461 (Education Week reports the OECD average fell by the same 14 points). Caveat: the U.S. weighted school response rate was 54% after replacement against an 85% target, per the NCES technical notes.
The source says: "The average reading score of U.S. 15-year-olds in 2025 (490) was lower than"
332, lowest in PISA history
America's weakest teenage readers scored lower in 2025 than in any PISA test since 2000.
National Center for Education Statistics (NCES) nces.ed.gov
More, and the source text
What it measures: U.S. 10th-percentile PISA reading score, 15-year-olds, 2025 (332) versus every prior cycle: 2022 (356), 2018 (361), 2015 (364), 2012 (378), 2009 (372), 2003 (361), 2000 (363). The U.S. also had the largest gap (312 points) between its top (90th percentile) and bottom (10th percentile) readers of any participating system.
The source says: "At the 10th percentile, the reading score in 2025 (332) was lower than in every preceding cycle"
35%
About 1 in 3 American 15-year-olds scored below the baseline level of math skill on PISA.
National Center for Education Statistics (NCES) nces.ed.gov
More, and the source text
What it measures: Share of U.S. 15-year-olds below PISA mathematics proficiency level 2 (baseline), 2025. The U.S. math average (463) was not significantly different from the OECD average (463) and was lower than in 2003 (483). Only 8% of U.S. students reached level 5 or above.
The source says: "35 percent were low performers in mathematics, scoring below proficiency level 2."
597 vs. 482 points
Students in four Chinese provinces scored more than 100 points above the OECD average in science, followed by Singapore, Macao, Chinese Taipei and Japan.
OECD (Education and Skills Today) oecdedutoday.com
More, and the source text
What it measures: PISA 2025 science mean scores, 15-year-olds: Beijing, Shanghai, Jiangsu and Zhejiang (China) 597; OECD average 482; Singapore 560; Macao (China) 541; Chinese Taipei 540; Japan 538. Science was the main PISA 2025 subject. The U.S. scored 502 in science (NCES).
The source says: "topped the PISA science test, achieving 597 points on average"
28-point drop in reading
Across OECD countries, 15-year-olds' reading scores fell 28 points over the past decade, and the OECD says 20 points is roughly a year of learning.
OECD (Education and Skills Today) oecdedutoday.com
More, and the source text
What it measures: PISA average reading score change for 15-year-olds across OECD countries, 2015 to 2025 (16 points in non-OECD systems). Same article: math fell 22 points across the OECD over the same period, and '20 score points is roughly equivalent to a year of learning gains'. Written by Andreas Schleicher, OECD Director for Education and Skills.
The source says: "28 points among OECD countries and 16 points among other countries and economies over the past decade"
29%, up from 25% in 2022
Nearly 3 in 10 teenagers in the world's wealthier countries are now low performers in reading, math and science, up from 1 in 4 three years earlier.
OECD (Education and Skills Today) oecdedutoday.com
More, and the source text
What it measures: Share of 15-year-olds in OECD countries estimated to be low performers across all three PISA subjects, PISA 2025 versus PISA 2022. The same article says the gap between low and high performers is now the largest ever recorded by PISA.
The source says: "some 29% of 15-year-olds in OECD countries are now estimated to be low performers across all three subjects"
22.6%
More than 1 in 5 American students missed at least 10% of the school year in 2024-25, compared with 15% before the pandemic.
American Enterprise Institute (Nat Malkus) aei.org
More, and the source text
What it measures: Chronic absenteeism (K-12 students missing 10% or more of the school year) in districts in 44 states plus DC with 2025 data, school year 2024-25. Peak was 28.3% in 2022; the national average was 15% in 2018 and 2019.
The source says: "these rates declined to 25.7 percent in 2023, to 23.5 percent in 2024, and to 22.6 percent in 2025."
411,549 positions (about 1 in 8)
About 1 in 8 U.S. teaching jobs is either empty or filled by someone not fully certified to teach that class.
Learning Policy Institute learningpolicyinstitute.org
More, and the source text
What it measures: Teaching positions unfilled (45,582 in 31 states plus DC) or held by teachers not fully certified for their assignments (365,967 in 48 states plus DC), 2025 state data compiled by LPI. Depends on state reporting and state certification rules.
The source says: "411,549 positions were either unfilled or filled by teachers not fully certified for their assignments."
28%, up from 19% in 2017
More than 1 in 4 American adults now read at the lowest level measured, up from about 1 in 5 just six years earlier.
National Center for Education Statistics (NCES) nces.ed.gov
More, and the source text
What it measures: Share of U.S. adults ages 16 to 65 scoring at or below Level 1 in literacy, PIAAC 2023 versus 2017. In numeracy the share rose from 29% to 34%.
The source says: "in literacy this percentage increased from 19 to 28 percent and in numeracy from 29 to 34 percent."
44 million teachers
The world needs 44 million more teachers by 2030 to put every child through primary and secondary school.
UN News (United Nations) news.un.org
More, and the source text
What it measures: Additional teachers needed by 2030 to reach universal primary and secondary education (SDG 4), from UNESCO's first Global Report on Teachers (2024). Sub-Saharan Africa alone needs an estimated 15 million. The same article reports primary teacher attrition rose from 4.62% in 2015 to 9.06% in 2022.
The source says: "The world urgently needs 44 million teachers by 2030"
273 million
273 million children, teens and young adults worldwide are out of school, and that number has gone up seven years in a row.
UNESCO Global Education Monitoring Report unesco.org
More, and the source text
What it measures: Global out-of-school children, adolescents and youth (primary through upper-secondary age), UNESCO 2026 GEM Report. UNESCO says this is one in six of that age group. The prior estimate (2025 SDG 4 Scorecard) was 272 million for 2023.
The source says: "273 million children, adolescents and youth are out of school, a rise for the seventh consecutive year."
70% of 10-year-olds
In low- and middle-income countries, about 7 in 10 children cannot read and understand a simple text by age 10.
World Bank worldbank.org
More, and the source text
What it measures: Learning poverty: share of 10-year-olds in low- and middle-income countries unable to read and understand a simple written text. 2022 simulation (World Bank, UNESCO, UNICEF and partners), up from 57% before the pandemic. A modeled estimate, not a new test. UNICEF's Digital Education Strategy 2025 to 2030 page still uses the same measure: 'two-thirds of 10-year-olds in low and middle-income countries are unable to read with comprehension'.
The source says: "with an estimated 70% of 10-year-olds unable to understand a simple written text"
39% of skills
Employers expect about 39% of the skills workers use today to change or become outdated by 2030.
UNLEASH (reporting the World Economic Forum Future of Jobs Report 2025) unleash.ai
More, and the source text
What it measures: Employer expectations of core skill disruption, 2025 to 2030, World Economic Forum Future of Jobs Report 2025 (survey of 1,000+ employers covering 14 million+ workers in 55 economies). Same article: if the workforce were 100 people, 59 would need training by 2030, and 11 of them would not get it soon. WEF pages and the report PDF blocked automated fetching; figure taken from trade press coverage of the report.
The source says: "39% of existing skills will face disruption or be outdated over the next five years"
170 million created, 92 million displaced
Employers expect 92 million jobs worldwide to disappear and 170 million new ones to appear by 2030.
UNLEASH (reporting the World Economic Forum Future of Jobs Report 2025) unleash.ai
More, and the source text
What it measures: World Economic Forum Future of Jobs Report 2025 projection of jobs created and displaced globally, 2025 to 2030 (net 78 million). Separately, CNN (January 8, 2025) reported that 41% of surveyed employers intend to downsize their workforce as AI automates certain tasks. WEF pages blocked automated fetching; figure taken from press coverage.
The source says: "170 million jobs will be created in the next five years, compared to 92 million displaced"
What AI does to learning
The studies so far, each with its design and its limits. Some are small or early; none of them settles the question alone.
Study
In one Harvard physics course, students who used a carefully designed AI tutor at home learned more, in less time, than students in a strong in-class lesson.
Design: Randomized controlled experiment with a crossover design (randomized at the level of peer-instruction groups of 2 to 3 students).
Scientific Reports (Nature Portfolio), full text via PubMed Central pmc.ncbi.nlm.nih.gov
Design, limits and the source text
Full design: Randomized controlled experiment with a crossover design (randomized at the level of peer-instruction groups of 2 to 3 students). N = 194 eligible students (233 enrolled), Physical Sciences 2 (introductory physics for the life sciences), Harvard University, Fall 2023. Custom GPT-4 tutor built on the same pedagogical best practices as the in-class lessons, with pre-written answers. Pre-test and post-test. Median AI time on task 49 minutes vs. a 60-minute class.
Limits: Authors state an AI tutor should not replace in-person teaching; instructors should avoid using AI where students are likely to use it as a crutch; they do not presume structured AI tutoring will always outperform in-class active learning in all contexts; gains may depend on several factors including that the tutor's accuracy relied on pre-written answers. One course, one university, short-term outcome.
The source says: "students learn significantly more in less time when using the AI tutor"
Study
In a World Bank trial in Nigeria, six weeks of teacher-guided after-school AI tutoring raised students' test scores by a meaningful amount.
Design: Randomized controlled trial;
The World Bank (abstract as indexed on RePEc/EconPapers; World Bank Open Knowledge page returned 403) econpapers.repec.org
Design, limits and the source text
Full design: Randomized controlled trial; six-week after-school program; first-year senior secondary students; Edo State, Nigeria; data collected May to July 2024 (World Bank Reproducible Research Repository). Teacher-supported use of Microsoft Copilot (GPT-4). Pen-and-paper assessment. Sample size not stated in the fetched abstract or repository page.
Limits: The abstract frames impact as conditional on tutoring designed and used properly. The authors' blog notes attendance was hampered by flooding, teacher strikes and after-school work, and lists open questions on long-term effects, other subjects and possible negative effects. Results reflect the full package (AI plus teacher support), not the chatbot alone. Working paper, not a peer-reviewed journal article.
The source says: "a significant improvement of 0.31 standard deviation on an assessment"
Study
When an AI tool coached human tutors in real time, their students mastered more math topics, with the biggest gains for students of weaker tutors.
Design: First randomized controlled trial of a Human-AI system in live tutoring, preregistered analysis plan;
arXiv (Stanford University authors) arxiv.org
Design, limits and the source text
Full design: First randomized controlled trial of a Human-AI system in live tutoring, preregistered analysis plan; 900 tutors and 1,800 K-12 students from historically under-served communities (Title I schools, Southern U.S. district). Outcome: mastery on exit tickets.
Limits: AI supports a human tutor; it is not a student-facing AI tutor. Tutors flagged suggestions that were not grade-level appropriate. J-PAL's summary lists a different sample (783 tutors, 1,013 students) and notes the study was not set up to assess effects on statewide end-of-year math test scores. Preprint (v1 2024-10-03, v2 2025-01-26, v3 2026-10-10).
The source says: "students of lower-rated tutors experienced the greatest benefit, improving mastery by 9 p.p."
Study
High school students who used an unrestricted ChatGPT-style tool for math practice did better while using it but worse on their own exam afterward; a tutor version with guardrails avoided most of that drop.
Design: Field experiment with nearly a thousand high school math students;
Proceedings of the National Academy of Sciences (abstract as indexed on RePEc/EconPapers) econpapers.repec.org
Design, limits and the source text
Full design: Field experiment with nearly a thousand high school math students; two GPT-4 tutors (GPT Base, a standard ChatGPT-style interface, and GPT Tutor, with prompts designed to safeguard learning) vs. a no-access control; outcomes on AI-assisted practice and on an unassisted exam. Setting: high schools in Turkey (per Knowledge at Wharton, 2024-08-27).
Limits: One subject (math), one country, one school system. Exam grades measured shortly after the intervention. Knowledge at Wharton reports students who used AI were overly optimistic about their own learning, and that AI output can contain false information. Full PNAS text returned 403; findings checked against the authors' abstract.
The source says: "17% reduction in grades for GPT Base"
Study
In a small MIT lab study, people who wrote essays with ChatGPT showed less brain connectivity on EEG and had more trouble quoting their own essays than people who wrote without tools.
Design: Lab study with EEG.
arXiv / MIT Media Lab arxiv.org
Design, limits and the source text
Full design: Lab study with EEG. 54 participants in Sessions 1 to 3, divided into three groups (LLM, Search Engine, Brain-only); 18 completed a fourth session with conditions switched. Essay-writing task over four months.
Limits: Preprint, not peer reviewed; the project page says all conclusions are to be treated with caution and as preliminary. Limited number of participants from a specific geographical area; one LLM, text only; essay task not split into subtasks; EEG spatial resolution limits; findings are context-dependent (essay writing in an educational setting). The authors ask media not to use words like "stupid", "dumb", "brain rot", "harm" or "damage" to describe the results; copy using this item must follow that request.
The source says: "Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels."
Survey
Knowledge workers who trusted AI more reported thinking less critically about its output, while those more confident in their own skills reported thinking more critically.
Design: Survey of 319 knowledge workers who use GenAI at work at least weekly;
Carnegie Mellon University and Microsoft Research (ACM CHI 2025; author-hosted copy) advait.org
Design, limits and the source text
Full design: Survey of 319 knowledge workers who use GenAI at work at least weekly; 936 first-hand examples of GenAI use in work tasks; self-reported critical thinking and effort. Presented at CHI 2025, Yokohama, April 26 to May 1, 2025.
Limits: Authors list: participants sometimes conflated reduced effort using GenAI with reduced critical thinking (possible inaccurate self-reporting); subjective confidence may not match objective expertise; English-only survey; sample biased toward younger, more tech-skilled workers; tools keep changing. Correlational: association, not cause. Adults, not students.
The source says: "higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking"
Study
The OECD's review finds general chatbots help students finish tasks but not necessarily learn, while tools built for teaching can help.
Design: OECD Digital Education Outlook 2026: international review of emerging evidence on generative AI in education.
OECD, summarized by the European Commission Digital Skills and Jobs Platform digital-skills-jobs.europa.eu
Design, limits and the source text
Full design: OECD Digital Education Outlook 2026: international review of emerging evidence on generative AI in education.
Limits: Fetched via the European Commission Digital Skills and Jobs Platform summary (updated 2026-02-03); the OECD pages returned 403, so exact OECD wording and release date not confirmed (URL path indicates January 2026). Evidence review, not a single new trial.
The source says: "successfully completing tasks with GenAI does not necessarily translate into learning"
Survey
About one in four U.S. teens said they had used ChatGPT for schoolwork in 2024, double the share a year earlier.
Design: Survey of 1,391 U.S.
Pew Research Center pewresearch.org
Design, limits and the source text
Full design: Survey of 1,391 U.S. teens ages 13 to 17, Sept. 18 to Oct. 10, 2024, through Ipsos KnowledgePanel (probability-based), weighted.
Limits: Self-reported; ChatGPT only (not other chatbots); superseded for broad chatbot use by Pew's February 2026 report.
The source says: "The share of teens who say they use ChatGPT for their schoolwork has risen to 26%"
Survey
More than half of U.S. teens now use AI chatbots for schoolwork, and one in ten say they use them for all or most of it.
Design: Survey of 1,458 U.S.
Pew Research Center pewresearch.org
Design, limits and the source text
Full design: Survey of 1,458 U.S. teens and their parents, online, Sept. 25 to Oct. 9, 2025.
Limits: Self-reported; covers any chatbot (ChatGPT, Copilot, Character.ai and others), so not directly comparable to the ChatGPT-only 2023 and 2024 figures; no ChatGPT-specific schoolwork figure given.
The source says: "One-in-ten teens say they do all or most of their schoolwork with chatbots' help."
Survey
Six in ten U.S. teens say students at their school use AI to cheat at least somewhat often.
Design: Survey of 1,458 U.S.
Pew Research Center pewresearch.org
Design, limits and the source text
Full design: Survey of 1,458 U.S. teens and their parents, online, Sept. 25 to Oct. 9, 2025.
Limits: Measures teens' perception of how common AI cheating is at their school, not measured cheating rates.
The source says: "A majority of teens (59%) think using AI to cheat is a regular occurrence at their school."
United States
Stanford HAI's 2026 AI Index reported that over 80% of U.S. high school and college students use AI for school-related tasks, while only half of middle and high schools have AI policies.
Stanford Institute for Human-Centered AI (HAI) hai.stanford.edu
More, and the source text
Details: Report takeaway: 'Formal education is lagging behind AI, but people are learning AI skills at every stage of life.' No comparable Chinese student figure was found in a reputable source.
The source says: "Over 80% of U.S. high school and college students now use AI for school-related tasks"
Kids and AI safety
What the surveys, the health agencies, lawmakers in both parties and the courts have on record. Lawsuits are allegations, stated as allegations.
Survey
About one in eight U.S. teens say they have turned to an AI chatbot for emotional support or advice.
Design: Survey of 1,458 U.S.
Pew Research Center pewresearch.org
Design, limits and the source text
Full design: Survey of 1,458 U.S. teens and their parents, online, Sept. 25 to Oct. 9, 2025.
Limits: Self-reported; does not measure frequency, outcomes or which chatbots were used for support.
The source says: "12% say they've used these tools to get emotional support or advice."
Survey
Nearly three in four U.S. teens have tried AI companion apps, and one in three use them for social or emotional connection.
Design: Nationally representative survey of 1,060 U.S.
Common Sense Media commonsensemedia.org
Design, limits and the source text
Full design: Nationally representative survey of 1,060 U.S. teens ages 13 to 17, April 30 to May 14, 2025, conducted by NORC at the University of Chicago (719 AmeriSpeak probability panel plus 341 Prodege nonprobability interviews, TrueNorth calibration); margin of error plus or minus 4.2 points. AI companions defined to respondents as digital friends or characters (for example Character.AI, Replika, or ChatGPT or Claude used as companions), not homework helpers.
Limits: Self-reported; mixed probability and nonprobability sample (cumulative response rate 10.3% for the probability panel); funded by Common Sense Media, an advocacy organization. Some teens may have interpreted the definition broadly.
The source says: "Seventy-two percent of teens have used AI companions at least once, and over half (52%) qualify as regular users"
Survey
A third of teen AI companion users have talked to a bot instead of a person about something serious, and a third have been made uncomfortable by one.
Design: Same NORC survey;
Common Sense Media commonsensemedia.org
Design, limits and the source text
Full design: Same NORC survey; base is teens who use AI companions (n = 758).
Limits: Self-reported; most uncomfortable incidents were infrequent (19% once or twice). Same sampling limits as the companion-use item.
The source says: "Among AI companion users, 33% have chosen to discuss important or serious matters with AI companions instead of real people"
Data
About one in three U.S. high school students reported persistent sadness or hopelessness in 2025, down from 2023 but still above 2015.
Design: CDC national Youth Risk Behavior Survey 2025 (school-based questionnaire, nationally representative sample of U.S.
Centers for Disease Control and Prevention cdc.gov
Design, limits and the source text
Full design: CDC national Youth Risk Behavior Survey 2025 (school-based questionnaire, nationally representative sample of U.S. high school students); 2-year change 2023 to 2025 and 10-year trend 2015 to 2025. Released in Youth Health in Focus: Findings From the National Youth Risk Behavior Survey 2015-2025.
Limits: Self-reported, cross-sectional survey; describes prevalence, not causes. CDC notes the report highlights select statistically significant differences. 2015 value read from the CDC chart. Sample size not stated on the fetched pages.
The source says: "persistent feelings of sadness or hopelessness, which improved from 40% to 33%"
Data
Fewer U.S. high school students reported serious thoughts of suicide in 2025 than in 2023, though levels are about where they were a decade ago.
Design: CDC national YRBS 2025;
Centers for Disease Control and Prevention cdc.gov
Design, limits and the source text
Full design: CDC national YRBS 2025; 2-year change and 10-year trend. The 14% figure is in the CDC press release text; plan and attempt figures are from the Mental Health and Suicide Risks page chart.
Limits: Self-reported, cross-sectional; prevalence, not causes; no link to any technology is stated. Safe messaging: any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988) and follow reportingonsuicide.org guidelines (no method details, no sensational wording).
The source says: "seriously considering suicide, which improved from 20% to 14%"
Data
Suicide remained the second leading cause of death for U.S. young people ages 10 to 24 in 2023.
Design: NCHS Data Brief No.
National Center for Health Statistics, CDC cdc.gov
Design, limits and the source text
Full design: NCHS Data Brief No. 541 (Garnett and Zehner), National Vital Statistics System mortality data.
Limits: One-year comparison; trends, not causes. Safe messaging: any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988) and follow reportingonsuicide.org guidelines.
The source says: "Suicide was the second leading cause of death in 2023 for people ages 10–14, 15–24, and 25–44."
Government action
In 2023 the U.S. Surgeon General said there was not enough evidence to conclude social media is safe enough for kids.
Design: Surgeon General's Advisory, Social Media and Youth Mental Health (May 23, 2023), as described on the federal SchoolSafety.gov resource page.
SchoolSafety.gov (U.S. federal government), describing HHS Office of the Surgeon General advisory schoolsafety.gov
Design, limits and the source text
Full design: Surgeon General's Advisory, Social Media and Youth Mental Health (May 23, 2023), as described on the federal SchoolSafety.gov resource page.
Limits: The advisory says the evidence is still developing. hhs.gov returned 403; wording taken from the SchoolSafety.gov description of the advisory, not the advisory PDF itself.
The source says: "states that we cannot conclude social media is sufficiently safe for children and adolescents"
Government action
The Federal Trade Commission ordered seven AI companies to explain how they protect children and teens who use their companion chatbots.
Design: Section 6(b) study orders (information requests).
Federal Trade Commission ftc.gov
Design, limits and the source text
Full design: Section 6(b) study orders (information requests).
Limits: A 6(b) study does not have a specific law enforcement purpose; it is not an enforcement action or a finding of wrongdoing.
The source says: "The Commission voted 3-0 to issue the 6(b) orders to the seven companies."
Government action
In September 2025, parents testified before a U.S. Senate subcommittee about harm they say AI chatbots did to their children.
Design: Congressional hearing record, S.
U.S. Government Publishing Office (GovInfo), Senate Committee on the Judiciary govinfo.gov
Design, limits and the source text
Full design: Congressional hearing record, S. Hrg. 119-256 (Serial No. J-119-39).
Limits: Testimony states witnesses' accounts and views, not adjudicated findings. Safe messaging: the hearing concerned children who died by suicide or self-harmed; any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988), follow reportingonsuicide.org guidelines, and include no method details from testimony.
The source says: "EXAMINING THE HARM OF AI CHATBOTS"
Government action
A bipartisan Senate bill would ban AI companions for minors and require chatbots to verify users' ages and say they are not human.
Design: Introduced bill text (119th Congress).
U.S. Government Publishing Office (GovInfo) govinfo.gov
Design, limits and the source text
Full design: Introduced bill text (119th Congress).
Limits: A bill, not law. congress.gov returned 403; status from the introduced text and the committee item below. Safe messaging: the bill addresses chatbots that promote suicide or self-harm; any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988).
The source says: "a covered entity shall prohibit the minor from accessing or using any AI companion"
Government action
In April 2026 the Senate Judiciary Committee unanimously advanced the bipartisan GUARD Act; it still needs votes in the full Senate and the House.
Design: Committee markup vote, reported by Roll Call.
Roll Call rollcall.com
Design, limits and the source text
Full design: Committee markup vote, reported by Roll Call.
Limits: Secondary news report; not yet law. Sen. Blumenthal said the committee vote is not the end of the road.
The source says: "voted 22-0 to approve the bill"
Government action
California became the first state to set safety rules for AI companion chatbots, including crisis referrals and protections for minors.
Design: State law (Senate Bill 243, Sen.
Office of California State Senator Steve Padilla (sd18.senate.ca.gov) sd18.senate.ca.gov
Design, limits and the source text
Full design: State law (Senate Bill 243, Sen. Steve Padilla), as described by the author's office.
Limits: Description from the bill author's press release; statute text at leginfo returned 403. Safe messaging: any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988) and follow reportingonsuicide.org guidelines.
The source says: "implement a protocol for addressing suicidal ideation, suicide, or self-harm"
Government action
California's governor signed one chatbot safety law but vetoed a stricter bill that would have barred most companion chatbots for kids.
Design: Governor's veto of a state bill, as reported by KQED.
KQED kqed.org
Design, limits and the source text
Full design: Governor's veto of a state bill, as reported by KQED.
Limits: Secondary news report; the veto message itself was not fetched. Safe messaging: the bill concerned chatbots encouraging self-harm; any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988).
The source says: "AI is already shaping the world, and it is imperative that adolescents learn how to safely interact with AI systems."
Government action
New York requires AI companion apps to spot signs of suicidal thinking, point users to crisis help, and remind them they are talking to a machine.
Design: State law, as described by the Governor's office (letter to AI companion companies).
Office of the Governor of New York governor.ny.gov
Design, limits and the source text
Full design: State law, as described by the Governor's office (letter to AI companion companies).
Limits: Applies to users generally, not only minors. Penalty amounts not stated on the page. Safe messaging: any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988) and follow reportingonsuicide.org guidelines.
The source says: "AI Companions must detect signs of suicidal ideation or self-harm"
Lawsuit (allegations)
California courts have grouped a dozen lawsuits against OpenAI over alleged ChatGPT harms into one coordinated proceeding.
Design: Court order (coordination of civil cases).
Superior Court of California, County of San Francisco (order copy hosted by Reason) reason.com
Design, limits and the source text
Full design: Court order (coordination of civil cases).
Limits: Coordination is procedural; no finding of liability. Several included cases involve adults, not minors. Safe messaging: any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988).
The source says: "CHATGPT PRODUCT LIABILITY CASES"
Lawsuit (allegations)
Character.AI and Google agreed in principle to settle several family lawsuits over alleged harm to teens, without admitting liability.
Design: Court filings (notices of resolution), as reported by Fortune;
Fortune fortune.com
Design, limits and the source text
Full design: Court filings (notices of resolution), as reported by Fortune; Decrypt reports the Garcia parties asked the court for a 90-day stay to finalize documents.
Limits: Settlement in principle; terms undisclosed; no admission of liability; allegations never adjudicated. Secondary source. Safe messaging: any page using this item must carry the 988 Suicide and Crisis Lifeline (call or text 988).
The source says: "no admission of liability appears in the filings"
What leaders on both sides say
Democrats and Republicans, alternating, in their own words, with the date and where they said it. Several of them wrote the same bill together.
Same bill: the GUARD Act (S.3062), introduced together on October 28, 2025
"In their race to the bottom, AI companies are pushing treacherous chatbots at kids and looking away when their products cause sexual abuse, or coerce them into self-harm or suicide. Our legislation imposes strict safeguards against exploitative or manipulative AI, backed by tough enforcement with criminal and civil penalties. Big Tech has betrayed any claim that we should trust companies to do the right thing on their own when they consistently put profit first ahead of child safety"
"AI chatbots pose a serious threat to our kids. More than seventy percent of American children are now using these AI products. Chatbots develop relationships with kids using fake empathy and are encouraging suicide. We in Congress have a moral duty to enact bright-line rules to prevent further harm from this new technology"
Same bill: the GUARD Act (S.3062), cosponsors
"You can’t overstate the harm these AI companion chatbots pose to our kids. Nearly 70 percent of teenagers are turning to chatbots as a substitute for friendship, making them more isolated and disconnected, but what’s even scarier are the stories of chatbots encouraging young people to hurt themselves or others."
"In recent months, we have seen multiple alarming incidents involving chatbots—from encouraging suicide to engaging in sensual conversations with minors—that should give every American pause. It is our job in Congress to do everything we can to protect our children as we navigate the new reality of living in a world where AI is increasingly integrated into our lives."
Same bill: the House GUARD Act, introduced together on April 30, 2026
"People under the age of 18 should not be able to interact with AI chatbots. These chatbots continue to put the lives and mental health of children at risk, and it is critical for Congress to act immediately"
"While our AI development agenda should seek to innovate and break barriers, it must also protect children from addictive and manipulative technology"
Same bill: the AI LEAD Act (S.2937), introduced together on September 29, 2025
"Democrats and Republicans don’t agree on much these days, but we’ve struck a remarkable bipartisan note in protecting children online. Big Tech’s time to police itself is over. Kids and adults across the country are turning to AI chatbots for advice and information, but greedy tech companies have designed these products to protect their own bottom line—not users’ safety."
"When a defective toy car breaks and injures a child, parents can sue the maker. Why should AI be treated any differently? This bipartisan legislation would apply products liability law to Big Tech’s AI, so parents—and any consumer—can sue when AI products harm them or their children"
Same bill: the NSF AI Education Act, introduced together on March 2, 2026
"Demand for AI expertise is already high and will continue to grow, and we must meet that demand if we are to out-compete China and make the American AI tech stack the global standard."
"As the capabilities of AI increase, we must make certain the next generation can utilize its full potential, supporting a strong economy, building thriving communities and maintaining America’s global leadership. Through investments in STEM education, fellowships and hands-on-learning from kindergarten to college, this legislation takes an all-of-the-above approach that will provide critical opportunities to students from all across the country."
"It’s alarming to see AI chatbots contributing to incidents of self-harm among young people"
"To ensure the United States remains a global leader in this technological revolution, we must provide our Nation’s youth with opportunities to cultivate the skills and understanding necessary to use and create the next generation of AI technology."
"Emerging technology like chatbots and social media can inspire, educate, and connect – but without real guardrails, technology can also exploit, mislead, and endanger our kids. We’ve seen some truly horrific and tragic examples of young people harmed by unregulated tech, and we won’t stand by while companies continue without necessary limits and accountability. We can continue to lead in AI and technology, but we must do it responsibly — protecting our children every step of the way. Our children’s safety is not for sale."
"Artificial intelligence has the potential to revolutionize education and support improved outcomes for learners"
"In the age of AI, it’s more important than ever that the people must govern. And as the land of liberty, America — not China — must lead the world on the development of AI."
"The country that leads in AI will shape the 21st century global order. As a matter of economic security, as a matter of national security, America has to beat China in the AI race."
What the smartest people say
The people building AI, studying it and teaching with it. Some warn, some hope, several do both.
"But I think we're at the cusp of using AI for probably the biggest positive transformation that education has ever seen. And the way we're going to do that is by giving every student on the planet an artificially intelligent but amazing personal tutor."
"The AI's will get to that ability, to be as good a tutor as any human ever could"
"We, as the producers of this technology, have a duty and an obligation to be honest about what is coming."
"Cancer is cured, the economy grows at 10% a year, the budget is balanced — and 20% of people don't have jobs"
"We prioritize safety ahead of privacy and freedom for teens; this is a new and powerful technology, and we believe minors need significant protection."
"It’s very hard to predict the future, like 10 years from now, in normal cases. It’s even harder today, given how fast AI is changing, even week by week"
"You're not going to lose your job to an AI, but you're going to lose your job to someone who uses AI."
"China is nanoseconds behind America in AI"
"Over the last year, a few people have been advising others to not learn to code on the basis that AI will automate coding"
"What really motivated me to put it out now before waiting for a full peer review is that I am afraid in 6-8 months, there will be some policymaker who decides, ‘let’s do GPT kindergarten.’ I think that would be absolutely bad and detrimental"
"The way we adapt is by preventing kids from having these friendships"
"AI holds tremendous promise but huge challenges—and it’s our job as educators to make sure AI serves our students and society, not the other way around"
"When we interview a software engineer, I personally feel the degree they have matters less to us now"
"Until it kind of exploded in my face thinking about my children, whether they would have a future."
What other countries' leaders say
Heads of government on putting AI in their schools.
"Our responsibility is to equip our children for a time unlike ours, with conditions different from ours, and with new skills and capabilities that ensure the continued momentum of development and progress in our nation for decades to come."
"That’s up to 1 million young people trained in tech skills—that will be so crucial for their future. That’s a £185 million investment, embedding AI right through our education system, starting in our secondary schools with subjects like computer science."
"one of the great worries for working people, and the same for any parent, is always: What does this mean for my children? What does it mean for their future?"
"Loss of jobs is AI's most feared disruption, but history has shown that work does not disappear due to technology, its nature changes and new types of jobs are created. We need to invest in skilling and re-skilling our people for an AI-driven future"
"The development of artificial intelligence should not be a solo performance by any single country but rather a symphony of global cooperation"
Education is our only priority.
Leave left and right at the door. Keep your position on everything else, and move this one forward together.