NYC Council meetings, chaptered and linked by time.
Created by Uzair Qadir
Inspired by Vikram Oberoi's City Meetings
New York City Council
Oversight - Artificial Intelligence, Student Data, and Privacy in NYC Public Schools
Read summaries and jump into the video at the segment start.
Council staff welcomes attendees to the joint Committee on Education and Technology hearing, reminds audience to silence devices and not approach the dais, and invites those wishing to testify to fill out appearance cards.
Councilmember Dinowitz introduces himself as Education Committee Chair and former teacher, framing the hearing around AI proliferation in schools, DOE's preliminary AI guidelines issued in May, and the intense debate over student-facing AI tools.
Dinowitz details calls from council members and CECs for a two-year moratorium on student-facing AI, cites a 2026 state comptroller audit finding DOE lacks a data classification policy, and references the PowerSchool breach exposing millions of students' data. He expresses concern about cognitive offloading and algorithmic bias.
Councilmember De La Rosa introduces herself as Technology Committee Chair and public school parent, outlining concerns about AI tools collecting student data including names, behavioral analytics, medical history, and facial geometry. She cites 141 DOE data breaches between January 2023 and February 2025.
De La Rosa warns that AI tools may profile students, surveil behavior, or make biased assumptions about abilities. She notes she joined colleagues in calling for a two-year moratorium and expresses disappointment that the Chief Privacy Officer and Office of Algorithmic Accountability are absent from the hearing.
De La Rosa thanks Technology Committee staff including Legislative Counsel Irene Bivolsky and policy analyst Eric Brown, and recognizes council members Schulman, Hanif, Gutierrez, Suong, and Wong in attendance.
Dinowitz introduces Intro 684 requiring a study on language access and geography affecting gifted and talented program access, and Intro 853 on language access, income, and geography affecting special education access, both sponsored by Councilmember Zhuang.
Zhuang argues that children in outer boroughs, lower-income families, and English language learners have less access to specialized programs. She shares examples of students traveling over two hours to special education schools and calls for expanding gifted and talented programs rather than eliminating them.
Dinowitz introduces his pre-considered bills: Intro 2129 requiring annual reporting on algorithmic tools approved for schools, and Intro 2130 requiring quarterly reporting on data breaches of algorithmic tools. He emphasizes transparency as critical for turning AI oversight from trust to accountability.
Dinowitz reminds witnesses to fill out witness slips, notes no votes will be taken, limits testimony to two minutes and council questions to five minutes, and acknowledges Councilmembers Linda Lee, Kevin Riley, and Farah Louis have joined.
Dinowitz administers the truth affirmation to the panel of DOE and administration witnesses, calling on each individually including Alicia Osborne-Reinhardt, Jeff Klein, Veronica Queller, Dennis Doyle, Tara Carrozza, Scott Strickland, Alan Chang, and others.
Danielle Junta introduces herself as First Deputy Chancellor, describes her oversight of superintendents and key offices, and notes she is joined by colleagues from data privacy, DIIT, early childhood, and inclusive learning. She describes regular community conversations about AI and screen time.
Junta explains that when Chancellor Samuels assumed his role in January, NYC Public Schools published AI guidance on March 25, launched an AI landing page, opened a 45-day feedback window, and began collecting input from educators, families, students, and community members across the city.
Junta details the March guidance requirements: AI tools using student data must pass the IRMA privacy review, student data cannot train AI models, human review of AI outputs is required, and AI cannot replace educator judgment in high-stakes decisions. Student use requires educator guidance and critical evaluation.
Junta reports 6,491 survey responses during the 45-day feedback period. Key takeaways include recognition of AI benefits with safeguards, student interest in AI literacy alongside over-reliance concerns, and educator distinction between teacher-facing and student-facing uses. Top concerns were cognitive development, mental health, and data privacy.
Junta outlines values driving next policies: centering developmental appropriateness across grade levels, distinguishing between unstructured student-facing AI and carefully reviewed instructional tools, and recognizing different considerations for assistive technology required by IEPs versus general AI tools.
Junta emphasizes student safety as essential, describes the IRMA privacy review process and December 2024 AI-specific standards added requiring vendor disclosure. She acknowledges families want broader questions answered about instructional purpose, accessibility, and bias, and commits to community engagement in technology governance.
Junta summarizes that March guidance established a foundation but feedback shows need for more detail and specificity. She states DOE is finalizing a stronger approach for the upcoming school year so educators and communities have clarity before the new year begins.
Junta addresses the proposed bills: DOE supports goals of 2129 on algorithmic tool reporting and notes existing Local Law 35 reporting; supports 2130 on data breach reporting and notes existing state law compliance; and on Intro 412 regarding admissions disclosure, notes much of the information is already publicly available.
Dinowitz asks if DOE has determined a specific age when students should not use large language models. Dr. Pate and Dr. Holla confirm there is no developmental framework for AI age bands, though early childhood has screen time policies of 30 minutes per week for 2K through pre-K.
Dinowitz questions why DOE approved AI vendors without empirical evidence of benefit to cognitive development, noting Gemini, Claude, and ChatGPT come pre-installed on student devices. He challenges DOE to provide peer-reviewed studies supporting these tools.
Scott Strickland clarifies that ChatGPT, Claude, and Gemini are approved for staff use only, with Gemini available to about 35 high schools by principal request. Dinowitz challenges this, noting teachers and principals report students accessing these tools in classrooms, and asks about VPN circumvention.
Dinowitz presses on whether students can access AI tools through approved vendor subscriptions like Canva, and raises VPNs as a way students bypass DOE filters. Strickland acknowledges VPNs may be used but insists students are supposed to use the DOE network. Dinowitz notes multiple principals report students using ChatGPT in class.
Dinowitz reads from Local Law 35 about an OpenGen AI teaching assistant tool that generates responses to student prompts, contradicting testimony that no student-facing AI is deployed. Strickland confirms it is piloted in District 28 with parental consent. Dinowitz calls students 'guinea pigs' and demands evidence of efficacy.
Dinowitz continues demanding peer-reviewed evidence for AI tools. Dr. Alan Chang cites meta-analyses and randomized control trials from journals including NPJ Science of Learning and PNAS, arguing structured AI tools with teacher guidance can support learning. Dinowitz counters that teacher guidance applies to any educational tool.
Dinowitz asks whether teachers are actually guiding AI use and for what educational purpose. Junta responds about tutoring programs and adaptive online tools, and acknowledges implementation challenges, saying DOE needs to understand how guidance is landing in schools and escalate policy violations.
Dinowitz shares concerns from multiple principals who feel they are 'building the plane as they're flying it' on AI. He draws parallels to credit recovery programs where students sit at computers with no teacher interaction, warning AI could worsen this without proper guardrails.
Dinowitz asks about DOE's position on AI detection tools given documented false positive rates disproportionately flagging multilingual learners and students with disabilities. Strickland confirms DOE has no AI detection tools approved. Dinowitz notes DOE does have plagiarism detection tools but no AI-specific detectors for disciplinary hearings.
Dinowitz asks what training and certification DOE provides teachers for AI tool use and detection of AI-generated content. Junta describes training tied to curriculum programs like HMH, Amira, and Zurn. Dinowitz presses on whether HMH includes required AI-specific training, which Junta admits it does not.
Dinowitz expresses confusion over conflicting testimony about whether student-facing AI is in classrooms. Dr. Pate acknowledges AI is new, describes a central AI guidance training course completed by over 8,300 staff, and mentions partnerships with UFT for teacher-facing professional development. Dinowitz counters that tools are in classrooms without full training or AI literacy curriculum in place.
De La Rosa asks if the 6,491 feedback responses will be made public; Junta commits to sharing them. De La Rosa then asks about GoGuardian monitoring software. Strickland confirms some schools purchase GoGuardian to monitor real-time student device activity in class, but a separate Beacon module is not approved. Monitoring does not extend to take-home devices.
De La Rosa raises concerns that DOE has declined to provide parents comprehensive lists of approved AI products. Junta says DOE is working with superintendents to gather information on what tools are in schools and encourages parents to ask their schools. De La Rosa asks about contract registration through Checkbook NYC.
Dinowitz asks about nondisclosure agreements with vendors. Dennis Doyle clarifies DOE has data processing agreements with confidentiality provisions but no NDAs preventing disclosure of contract existence. Junta confirms all contracts are registered in Checkbook NYC and with the Comptroller.
Dinowitz asks about the AI playbook planned for June 2026 per DOE's guidance. Junta explains the original plan has shifted due to feedback and the evolving national conversation, saying DOE is not racing into a playbook but carefully building out stronger guidance. Dr. Pate adds the delay is about being responsive to feedback.
Dinowitz asks about YouTube access on DOE devices; Strickland says YouTube is generally blocked unless a principal requests it for educational purposes. Dinowitz then asks about Play Prodigy, an app that prompts students to pay for features. Strickland confirms the free version is approved but the pay features are not encouraged.
Dinowitz notes that 6,491 responses represents less than 1% of parents, calling it a very low percentage compared to the parent survey which has significantly higher response rates. Junta agrees and says DOE supplemented with targeted conversations with advocates, labor, parent leadership, and elected officials.
Zhuang expresses disappointment that DOE was unprepared to answer questions about her bills on G&T and special education access. She asks about the effect of G&T participation on specialized high school admission and cites her own research showing 12% higher admission chances. DOE officials say they do not have the data readily available.
Zhuang continues pressing for data on G&T program effects on later life outcomes and current program distribution in black and brown communities. DOE officials acknowledge they did not bring G&T-specific data. Suzanne Sanchez introduces herself as Chief of Special Education. Dinowitz instructs DOE to provide answers before leaving the hearing.
Councilmember Avilés cites the May 2026 state comptroller audit finding DOE lacks written data classification policies, risk assessment, and backup/recovery procedures required by the NIST framework under EdLaw 2D. Dennis Doyle responds that DOE complies with privacy laws, accepted audit recommendations, and is implementing improvements including a formal data classification policy.
Avilés asks when updated security protocols will be released. Doyle says data privacy and security policy updates will come in coming months, and notes DOE updated Chancellor's Regulation A20 on student privacy for the first time since 2009, incorporating feedback from parents, advocates, and school leaders.
Councilmember Wong asks where parent feedback on AI guidance is publicly available; Junta says it will be made public. Wong then asks about the June 2026 playbook timeline. Junta reiterates the playbook is shifting due to feedback and the national conversation, and DOE is building out stronger guidance instead.
Wong asks about policies for teachers using AI to plan lessons or grade papers, comparing it to HR departments using AI to screen resumes. Dr. Pate states assessments and grading are prohibited AI uses, and DOE does not recommend teachers use AI to generate lessons, emphasizing educator input must be culturally appropriate and accurate.
Councilmember Hanif asks if any biometric recognition technologies are deployed in schools. Strickland confirms no biometric detection tools are used and that such tools are prohibited by law. Hanif then asks how AI tool necessity is evaluated before introduction.
Dr. Pate describes a tool scorecard being developed to assess educational value and student outcomes, and beta testing with 50 teachers across grades 6-12 in core content areas. Tara Carozza adds that DOE is currently only evaluating teacher-facing tools, not direct student AI use, due to concerns from the feedback window.
Hanif asks if a similar scorecard exists for students. Carozza says DOE is creating evidence-based scorecards for all stakeholders including families, aiming for transparency in evaluation criteria. Hanif recommends including students in the evaluation process.
Councilmember Gutiérrez, a moratorium signatory, criticizes DOE's process as backwards—releasing guidance before collecting feedback—and says principals feel forced into using AI tools without preparation. She demands a digital literacy vision and desired outcomes from DOE before guidance or playbooks are issued.
Gutiérrez asks whether the original March AI guidance could be tossed after the playbook process. Junta says DOE stands by the March guidance but acknowledges the need to build more detail and specificity. Gutiérrez expresses frustration with the fragmented implementation approach.
Councilmember Joseph asks where parents are in the AI conversation, whether they are made aware of AI in classrooms, and if they can opt in or opt out. She also asks about language access for non-English-speaking parents and who the two leads are on AI digital learning. Dr. Pate describes superintendent-level parent engagement.
Joseph asks for a timeline to ensure every educator receives AI training, noting some students are 'running circles around their teachers' with AI. Dr. Pate acknowledges no timeline is set yet and agrees technology moves faster than training development, describing it as a challenge of chasing capacity building.
Joseph asks how DOE addresses biases in existing AI tools. Dr. Pate acknowledges there is no endorsed tool that tracks bias, calling it a 'mega problem,' and says DOE is looking for resources. She emphasizes AI literacy development and awareness that AI can be wrong as the primary approach.
Joseph asks about supporting English language learners and students with disabilities. Suzanne Sanchez, Chief of Special Education, describes the irrefutable evidence base for assistive technology for students with disabilities, noting it follows careful IEP evaluation processes and goes through IRMA vetting. Dinowitz clarifies this is distinct from unvetted AI tools in classrooms.
Dinowitz asks why DOE does not have a required digital literacy curriculum given decades of social media and media bias concerns. He argues DOE should not wait for state guidance and should ensure digital literacy and civics education happen in all schools, not just provide opportunities.
Dr. Chang responds that state law requires media literacy and that NY State Ed is updating guidance to include AI. He describes Future Ready NYC partnerships with industry and higher education, and emphasizes civic readiness as essential. Dinowitz pushes back that DOE should not wait for the state and should set expectations for all schools.
Councilmember Maloney asks what changes are expected in the revised AI framework and when it will be released, noting distinctions needed between teacher and student use cases and age appropriateness. Junta says DOE is working furiously but cannot commit to a date, and agrees agile governance with feedback loops is needed.
Maloney asks what system exists for continuously revisiting AI guidelines given the technology's rapid evolution. Junta agrees the iterative process with feedback loops must be baked into the next guidance evolution, and DOE officials emphasize ongoing engagement with educators, parents, and students.
Councilmember Krishnan asks how DOE teaches students to use AI responsibly, critically, and ethically—not as a replacement but as a tool to interrogate. DOE officials describe AI literacy as essential, and Tara Carozza details a 15-lesson high school module on AI concepts and ethics, with K-8 support planned for the coming year.
Krishnan emphasizes that high school graduates will be expected to know how to interact with AI tools in the workforce, and urges the next guidance to include specifics on age-appropriate AI civics. He acknowledges the challenge of building the plane while flying it but stresses the need to move faster.
Avilés asks why a vendor for the MOSL teacher evaluation algorithmic tool is listed as 'vendor not disclosed' in the Local Law 35 report. DOE officials explain it is an algorithmic tool predating generative AI and say they will follow up on whether it was developed internally or by a third party.
Avilés asks about the Chief Privacy Officer's role at OTI, whether DOE uses master service agreements, and whether OTI's citywide AI guidance binds DOE. Doyle explains DOE works closely with OTI on IRMA and cloud reviews but is subject to separate laws under FERPA and EdLaw 2D rather than the city's identifying information law.
Avilés asks how many privacy complaints DOE receives annually and typical outcomes. Doyle says he does not have specific numbers but describes the investigation and notification process. Avilés then asks about parental consent rights under the Parents' Bill of Rights for Data Privacy and Security.
Avilés asks how parents can exercise the right to require written consent before education records are disclosed. Doyle explains DOE has ready-made consent forms on its website, and describes FERPA exceptions allowing disclosure without consent, such as to school officials with legitimate educational interest or in health emergencies.
De La Rosa questions Doyle about the state comptroller audit finding that only 73% of employees completed required privacy training in 2024, and that DOE could not verify which employees with access to student data had completed training. Doyle reports 138,000 of approximately 150,000-160,000 employees completed training in the most recent year.
De La Rosa asks about confidential weaknesses in system monitoring, unsupported systems, and firewalls identified in the comptroller's audit. DOE officials confirm changes have been made or are planned but decline to disclose specifics for security reasons. De La Rosa presses for status updates on each recommendation.
Wong asks specifically which AI tools principals, teachers, and administrators are using. DOE confirms ChatGPT, Gemini, and Copilot are used by staff, while Perplexity, Claude, and Consensus are not widely used. Wong asks about licensing and whether anyone has evaluated the quality and accuracy of different AI engines.
Wong presses on whether DOE has compared quality of results across different AI engines and whether commercial licenses are needed. DOE officials say Microsoft and Google products have free and paid versions, both approved, and emphasize human judgment and fact-checking are expected with all AI use.
Hanif raises concerns that Amira, a reading support tool, collects biometric voice data from students. Tara Carozza explains Amira uses student voice interactively for reading fluency practice, and under the data processing agreement, all data must be deleted after one year, hosted in the U.S., and not used to train other models.
Hanif asks whether reading comprehension is improving with Amira and requests evaluation data. Dr. Pate says longitudinal data is not yet available but shares that students using Amira up to three times weekly for 10 minutes showed about 4-5 months acceleration in reading comprehension compared to non-users, based on last school year's K-2 data.
De La Rosa follows up on Hanif's questions, asking why data processing agreements are not shared publicly. Doyle says it is DOE policy not to post internal records. De La Rosa argues it is a choice and says the Council expects to receive the agreements, noting no NDAs prevent disclosure per earlier testimony.
Zhuang asks DOE for data on special needs students by zip code, borough, race, and language access. Suzanne Sanchez says much of this data is publicly reported in City Council special education reports and offers to review it outside the hearing. Zhuang notes 75% of students with disabilities are black or Latino.
Zhuang cites data showing 71% of private school tuition reimbursement for special education went to white families, while only 12.5% of students with disabilities are white. Sanchez acknowledges the equity concern and describes efforts including inclusive districts in underserved neighborhoods, 900 intervention teachers, and expanded assistive technology access.
Zhuang shares constituent stories of immigrant families traveling 1.5 hours to Queens for special needs schools, and asks about data on students facing long commutes. Sanchez confirms DOE has District 75 data and is expanding specialized programs closer to home to reduce busing burdens. Zhuang also raises language access barriers for immigrant families.
De La Rosa reads from DOE's website stating IRMA does not evaluate algorithmic bias, then reads another section stating the legal team reviews bias. Doyle confirms the legal team does not review bias, calling it a possible website error. De La Rosa instructs DOE to correct the website, which has been public since March.
De La Rosa asks if DOE can commit to ensuring bias and equity impact are part of tool evaluation. Tara Carozza describes an equity and ethics fellowship and collaboration with OTI on bias evaluation, but acknowledges no solid tool currently exists for algorithmic bias assessment. De La Rosa calls the minimum standards 'shocking.'
De La Rosa cites the comptroller's finding that DOE reported 67 of 141 breaches late, nearly half. Doyle explains difficulties with vendor reporting timeliness, staff awareness, and forensic investigation delays. De La Rosa asks what steps beyond training have been taken to fix lapses in notifying families within the required 60-day timeframe.
De La Rosa details the Illuminate breach where 380,000 families were not informed for over two years. Doyle explains the timeline of forensic investigations and notifications. De La Rosa asks if DOE drops vendors who fail to report breaches; Doyle confirms DOE stopped doing business with Illuminate Education after the breach.
De La Rosa asks if DOE has included penalties in vendor contracts for data breach reporting failures. Doyle says DOE is looking into strengthening data processing agreements to include penalties for not notifying in a timely fashion. De La Rosa calls penalties 'incredibly appropriate' given the scale of student data involved with AI.
Gutiérrez asks what compliance standards are referenced in the IRMA security review and who sets them. DOE officials explain the three-step IRMA process: data processing review, DIIT technology review, and OTI cloud review, with standards set by DOE.
Gutiérrez asks about the prohibition on behavioral monitoring, noting the guidance says it's prohibited but also says DOE is reviewing how policies apply. Strickland confirms behavioral monitoring is in the red/prohibited area and DOE is ensuring it stays closed, not opening it up. Gutiérrez confirms the prohibition extends to device monitoring, emotional recognition, facial recognition, and profiling.
De La Rosa thanks the administration, summarizes key findings including the need for vendor penalties, tightened reporting transparency, and her two bills on algorithmic tool reporting and data breach reporting. She emphasizes the opportunity to embed digital literacy and civics through Portrait of the Graduate without waiting for state guidance.
De La Rosa opens the public testimony portion, reminding attendees of decorum rules, two-minute time limits, no video recording from the witness table, and procedures for submitting written testimony to testimony@council.nyc.gov within 72 hours.
Yolani Joseph, a 16-year-old sophomore and NYCLU Teen Activist Project organizer, shares an experience where a teacher's AI-generated reproductive health slideshow was full of errors. She calls for AI literacy beyond filling classrooms with products, demanding accountability, clear standards, and human judgment at the center of education.
Odin Adeller, an 18-year-old recent graduate, describes how his school's restriction on AI helped him develop writing confidence and critical thinking skills. He opposes DOE guidance encouraging AI for research and creative projects, and raises concerns about students being forced to sign away digital rights to big tech companies.
Arunamid Misha, a rising junior, describes student opposition to AI policies from a weekend event with hundreds of students. She argues AI takes away critical thinking skills, that students want human teachers not AI bots for feedback, and that AI bots create unrealistic perfectionist expectations that destroy student confidence.
Selena Wu, a rising senior who has built machine learning models at Princeton and MIT, opposes Gen AI in schools. She describes how an English teacher's AI grading incorrectly deducted points and falsely accused classmates of using AI. She argues AI makes learning more boring and calls for engaging learning environments instead.
De La Rosa asks if the students' views against AI represent most classmates. Students confirm that peers who care about their future oppose AI, noting that while some use it for convenience, students understand it impacts learning and the environment. Odin notes AI use is uncommon at his consortium school where class discussions demand original thinking.
Vinny Dahl, a 15-year-old sophomore, testifies about Intro 684 and geographic barriers to education, then demonstrates how AI detection tools fail by showing a ChatGPT paragraph that bypasses detectors and Chancellor Samuels' speech flagged as 40% AI. He argues DOE cannot effectively regulate AI and detection will harm innocent students.
Nasser-Frederick from the Collaborative Research Center for Resilience and the AI Moratorium Coalition warns that most ed tech vendors are designated as 'school officials' under FERPA, allowing data sharing without parental consent. She raises concerns about HMH's ownership by Veritas Capital, which also builds DHS biometric databases, and calls data processing agreements a cause for alarm.
Edgar Trout suggests distinguishing between top-down ed tech rollouts (where a pause is prudent), bottom-up teacher innovations, learning about AI itself through unplugged activities, and teachers using AI for lesson development. He argues beneficial uses should survive a moratorium and stresses partnership with higher education institutions.
Travis Proof from Fordham University outlines four pillars for AI in education: understanding AI foundations, discerning ethical versus unethical uses, augmenting rather than replacing work, and recognizing when AI is inappropriate. He describes Fordham's ethical AI hub and calls for a citywide training strategy beyond education.
De La Rosa asks Dr. Trout about learning AI without computers; he confirms unplugged activities are a computer science education tenet. De La Rosa asks Nasser-Frederick to clarify the FERPA school official loophole, which she explains has allowed vendors to access student data without consent since the 1970s.
Lucero testifies about Urban Assembly's AI-supported instructional feedback tools showing measurable improvements including 14% increase in student talk time and 14% decrease in classroom silence. She emphasizes AI works best when human-guided and student privacy is non-negotiable, noting recognition from the EdSafe AI Alliance.
Lieberman describes MOUSE's 30-year history training NYC public school students and teachers in technology, including an AI League at 40 schools with Carnegie Mellon curriculum. He reports 83 teachers attended a generative AI fluency training in one council district, calling the demand for AI training the largest in the organization's history.
Messier from New York Lawyers for the Public Interest supports Intro 853's goals but questions whether the study is the most effective use of resources, noting existing data already shows language, income, and geography affect access. She cites that only 28% of ELL students with IEPs receive services in another language and 47% of Spanish-speaking students receive recommended services versus 95% of English speakers.
Henning expresses TechNYC's support for the reporting bills but offers recommendations: clarify the broad definition of 'algorithmic tool' in 2129, coordinate with existing IRMA and Office of Algorithmic Accountability frameworks to avoid duplication, and work with DOE on the October 31st reporting deadline infrastructure for 2130.
De La Rosa asks Lucero whether using the EdSafe AI Alliance framework indicates DOE hasn't provided robust enough support. Lucero explains the tool has been recognized by EdSafe and there have been conversations with NYC DOE and OSP, though the exact relationship is unclear. De La Rosa notes it appears to be another evaluative tool outside DOE's framework.
Xu testifies in support of Intro 684, noting G&T programs are unevenly distributed with too few seats—10,000 families apply for 2,500 spots. He criticizes the sole reliance on subjective teacher recommendations for G&T identification, arguing it disproportionately harms immigrant families and ELL students, and calls for universal screening with standardized assessments.
Blommenbill testifies about being an unwitting victim of neurological attacks and technology weaponization, describing organized stalking, electronic assault, and non-consensual human experimentation. She thanks the Council for laws like the POST Act and biometric access restrictions.
Ritchie, a parent and former teacher, calls for an AI moratorium and reduction of ed tech products. She cites a New York Times article where Google said investing in schools helps onboard kids to its ecosystem, and argues schools are force-feeding tech products that collect and sell children's private information. She calls for nurturing human connection in education.
Zam from Homecrest Community Services testifies in support of Intros 684 and 853, describing how immigrant families in southern Brooklyn face barriers including lack of translated materials, transportation challenges, and uneven program availability. She argues a child's opportunity should not depend on whether a parent speaks English or lives in a resource-rich neighborhood.
Senec from the Center for Independence of the Disabled New York warns that AI could mask critical symptoms and delay diagnoses for students with disabilities. She identifies risks including critical thinking erosion, bias, misinformation, mental health impacts, and trust erosion, and calls for getting ahead of AI regulation before problems seen with cell phones and social media recur.
Walker, a parent of two DOE device users, contradicts DOE testimony that YouTube is not on devices, saying her children have full internet and YouTube access. She describes AI-powered Amira collecting children's voices without parental consent and calls for transparency about what technology children interact with at school.
Adedon, a freelance creative professional and parent, urges a moratorium on predatory LLMs, noting they are built on stolen work of writers and artists. They cite APA health warnings about adolescents and AI, documentation of chatbots inducing psychosis in adults and driving children to suicide, and argue AI will stunt critical growth for thousands of students.
Stowe, a former CEC member, supports Intros 684 and 853, citing Chalkbeat data that 71% of private school tuition for students with disabilities went to white families who make up only 12.5% of students with disabilities. He calls for addressing awareness gaps and expanding G&T programs in underserved districts rather than making token efforts.
Zachary, former PEP representative, expresses dismay that privacy suggestions from years of task forces were ignored. She calls a two-year moratorium almost laughable given DOE's slowness, and argues AI is inevitable—the focus should be on safety measures, privacy, and integrity guardrails rather than avoidance.
Comic book writer Greg Pak calls AI in schools 'child abuse,' arguing that writing is thinking and drawing is seeing. He says kids who type prompts instead of doing the work themselves are denied human learning experiences. He calls AI a tool of liars and scammers and warns billionaires can push a 'racism button' on these machines.
Rebellini, a writer and mother, shares stories of her father's 43-year teaching career and the immeasurable impact teachers have that machines cannot replicate. She cites her own book about suicide to emphasize the consequences of losing human connection, noting black children are especially susceptible to isolation, and calls chatbots used as study guides harmful.
Gad urges redirecting funds from tech companies and surveillance tools toward organizations supporting young people's tech literacy and safety infrastructure including restorative justice and mental health access. She describes a youth town hall where students demanded best practices for AI, data privacy education, and consentful tech.
Pope from the NYCLU focuses on the student surveillance industry, noting that FOILed data processing agreements are heavily redacted under trade secrets exemptions. She describes AI products that monitor devices, track physical movements and social connections, and assign mental health wellness scores, calling them a digitalized school-to-prison pipeline and questioning DOE's enforcement of prohibitions.
Andrea testifies about excessive policing and surveillance in schools, noting black students make up less than 20% of enrollment but 45% of superintendent suspensions and 52% of police interventions. She calls for redirecting $90 million from school safety to restorative justice and mental health, and enacting a two-year AI moratorium.
Ramirez, a CEC District 4 member and neuroscientist, calls for a two-year AI moratorium, reduced screen time, and an audit of iReady, which she calls an ineffective tool costing over $100 million. She cites Dr. Jared Horvath's research showing ed tech has harmed students and argues adolescence now extends to age 30 due to lack of experiential learning.
Stoyanovich, NYU professor and founding director of the Center for Responsible AI, makes three asks: prove AI tools work through independent efficacy and equity standards before classroom use, keep student data home rather than flowing to vendor platforms, and open data for independent verification. She warns AI tutors are flooding fastest into low-income schools.
Irfan, founder of Augmented Intelligence Advisory, cites Microsoft and Carnegie Mellon studies showing AI reliance reduces critical thinking. She warns of a cognitive divide along socioeconomic lines and urges deploying cognition-focused AI literacy where students form hypotheses before using chatbots, and adding a cognition check to Bill 2129 requiring evidence of a tool's effect on student reasoning.
Kazi, a parent in District 6, describes a 'break in communications' between DOE and classrooms, saying her fifth grader used Canva AI and YouTube in class and her third grader is prompted by Google Gemini in Google Classroom. She cites a Columbia Teachers College study favoring textbooks and paper, and calls iReady—which has AI and is being sued—an equity issue.
Mariela, a parent of a child entering kindergarten, expresses fear about AI in classrooms, saying it is 'almost negligent' to put chatbots in front of the most vulnerable population. She describes parents working hard to limit technology at home only to have schools introduce AI, and calls for immediate action before summer.
Thacker, an NYU English professor and parent, calls the AI guidance sloppy and lacking any evidence-based rationale for AI as a pedagogical good. She argues the guidance represents capitulation to a contested technology, calls children 'not guinea pigs,' and warns that parents with means will exit public schools for those with restricted AI policies.
Hatterer, an education technology policy consultant, opposes the two-year moratorium, citing Pew Research data that 51% of adults use AI chatbots and 95% of teens have heard of AI. She argues avoidance will diminish student agency and leave them less prepared, and references a Brookings Institute three-pillar strategy of Prosper, Prepare, and Protect rather than a pause.
Monahan from Fair Play describes a position statement signed by 260 organizations and experts calling for a five-year pause on Gen AI products for students. He lists five requirements for reintroduction: improved learning without cognitive offloading, absolute safety, no cheating, privacy and civil rights prioritization, and never replacing teachers for vulnerable populations.
Dwyer from the Center for Democracy and Technology identifies four key AI risks from national research: data breaches, deepfakes, AI system failures, and troubling student-chatbot interactions. She recommends risk management, AI governance, transparency requirements for vendors, model procurement resources, and formal avenues to engage parents and students in AI decisions.
East, a parent in De La Rosa's district, shares her child's experience with Zurn and IXL, describing how these tools took her struggling learner back to lower levels rather than identifying the problem. She criticizes IXL's 'smart score' as devastating to children's self-esteem, contrasting AI tools with human teachers who can pivot approaches.
Rao, a Citywide Council on Special Education member, testifies about the lack of adherence to federal special education law and the long-standing use of AI in assistive tech without guardrails. She asks why DOE doesn't mandate GoGuardian across all schools with a central budget rather than leaving it to individual school choice.
Senda, a parent at PS 184, describes his son looking forward to classes because of games dressed up as education, warning of addictive technology attachment. He notes a power imbalance between DOE and the Council, where the Council seems to be treated like an outside entity, and asks how parents can support the Council's oversight efforts.
Carmen, mother of two NYC public school graduates, testifies that Gen Z expresses distrust and alarm at AI's encroachment. She shares her daughter's experience with an AI job portal that misrepresented her resume, and her daughter's decision to stop using AI to avoid losing her ability to write. She warns AI in schools serves standardized testing obsession.
Dinowitz thanks co-chair De La Rosa and all participants, summarizing that the hearing revealed significant gaps in DOE's AI policy including bias and equity evaluation, student safety, academic impact, and cognitive offloading. He commits to pushing for answers from DOE and shaping policy that keeps students safe while maintaining rigorous academic standards. The hearing is adjourned.