Five Minutes on AI Agents in Education: Google Will Turn Off School AI by Default, but Ties It to Search

This week, Google said schools that restrict its AI tools for students altogether will also lose Google Search.¹ Here are five stories and what they mean for educators.

Google's new K-12 defaults: AI off, and Search tied to it

What happened. Google says Gemini, Gemini Notebook, Gemini in Classroom, and Search will be off for K-12 student accounts until administrators turn them on. It says it will share details for schools "in the coming months."² Education Week reports that the changes start early next year. Administrators can switch on Gemini, Search, or both, and can set different rules by grade. Any group barred from AI also loses Search.³

Sen. Bernie Sanders, Sen. Richard Blumenthal, former FTC chair Lina Khan, and New York City Mayor Zohran Mamdani criticized the plan.⁴ Critics cited by K-12 Dive say schools must either accept Search with its AI features or lose Search. The story also reports that New York City Public Schools and Los Angeles Unified recently paused student-facing AI tools.⁵

The strongest view on the other side. After the criticism, Google said Gemini, Gemini Notebook, and Search are off by default and that schools can turn each one on individually.⁴ On that view, the choice now sits with administrators.³ That is a fair point. But it does not help a district that wants Search without AI, because any group barred from AI also loses Search.³

What it means for you. GovTech reported on October 26 that Sens. Blumenthal and Hawley had asked Google's CEO how the company answers concerns from districts, teachers, and parents.⁶ If you lead a school or district, your own teachers and families may ask you the same kind of question about your choice. Decide now who will make the call for each tool and grade, and how you will explain it.

UIC asks faculty to disclose their own AI use

What happened. The University of Illinois Chicago's AI Use and Disclosure Statement requires faculty to disclose the AI use they plan in course design, instructional materials, assessment, and grading support. Each syllabus must also say which student uses of AI are acceptable and which are prohibited.⁷

What it means for you. UIC's statement covers faculty as well as students. If you ask students to be open about their AI use, consider whether you are as open with them about yours.

A study finds AI peer reviews rated above human ones

What happened. Harvard researchers compared 134 human peer reviews of 124 papers with reviews written by Anthropic's Claude Opus. When three former journal editors judged the pairs, the AI review beat the human review in 87% of them. The paper has not yet been peer reviewed.⁸

What it means for you. This is one study, and it has not been peer reviewed, so treat it with caution. Still, it gives faculty and journal editors a reason to discuss where AI belongs in reviewing scholarly work.

A district builds an AI tool for bus routing

What happened. Education Week reports that Sweet Home Central School District near Buffalo, which serves about 3,800 students, built a Claude-based tool to route after-school buses. The job takes staff about six hours per run. The district also plans to bring AI into one administrative department, which it has not yet chosen.⁹

What it means for you. This is our hypothesis, drawn from one district's example: a well-defined, time-consuming administrative job may be a sensible first place to try AI in a school, because the time saved can be measured. Evidence that such tools save little staff time or add new errors would change our mind.

A ransomware attack on UK schools, and how much AI was involved

What happened. Bitdefender investigated a ransomware attack on a UK school estate that ran from September 10 to 19. About 37 of about 50 reachable machines were encrypted. Bitdefender rates the use of AI-written commands as strongly supported by the evidence.¹⁰ An Elon University article defines agentic AI as AI tools that can carry out multistep tasks with some independence.¹¹ Bitdefender says an AI agent running the attack is only a model for defenders to plan around, not an established fact, and it calls the case "an anecdote, and not a trend."¹²

What it means for you. Bitdefender keeps what its evidence shows separate from what defenders should plan for. You can do the same. Ask your IT team what, if anything, in your plans would change if attackers used AI agents, without treating one case as a trend.

One question for this week

If your district had to choose tomorrow whether students get Google's AI tools, Search, both, or neither, who would decide, and how would you explain the decision to your teachers and families?