New AI tool logs into Canvas to complete homework automatically as 54% of teens use chatbots

Einstein AI Automates Coursework Through Canvas Integration

A new AI tool called Einstein, created by startup Companion.AI, has pushed automation in education into uncharted territory. Unlike AI chatbots that respond to prompts, Einstein AI operates as a digital stand-in that logs directly into student accounts on Canvas, a widely used learning management system in colleges and high schools 1

1

. The system can watch lecture videos, read PDFs and essays, write papers, complete quizzes, post replies in discussion boards, and automatically submit assignments before deadlines 5

5

.

According to Advait Paliwal, CEO of Companion.AI and a Brown University dropout, Einstein operates through its own virtual computer with browser automation capabilities. "Our companions aren't simple chatbots," Paliwal said. "Each one has access to an entire virtual computer with a persistent file system and internet access, so they can actually do things on your behalf. This makes ChatGPT look like a toy" 1

1

. The technology builds on advances in generative AI and autonomous agents that can take multistep actions without ongoing human input.

Majority of Students Already Using AI for Homework

The release of Einstein comes as new data from the Pew Research Center reveals that 54% of U.S. teens aged 13 to 17 have used AI tools for homework 4

4

. The survey, collected in fall 2025, shows that 10% of all respondents report using AI for all or most of their homework, while 44% use it for a little or some coursework 3

3

. Only 45% of students haven't used AI chatbots for school assignments at all.

Roughly four in ten teens use AI to research topics or solve math problems, while a third have used it to edit their writing 4

4

. About a quarter of all teens say these tools have been extremely or very helpful for schoolwork, with another quarter finding them somewhat helpful. The positive experience keeps adoption climbing as schools scramble to develop policies addressing AI use.

Disparities Emerge in AI Dependence Patterns

The Pew data reveals troubling disparities in how students rely on AI. Twenty percent of kids in households making less than $30,000 a year reported doing all or most of their homework with AI's help, compared to just 7% of kids whose households earn over $75,000 3

3

. Black and Hispanic teens are also 12% more likely than their white counterparts to do all or most of their schoolwork with AI chatbots.

These findings emerge as the federal share of K-12 education funding continues a 50-year decline, suggesting that minority and low-income students may be turning to AI tools to compensate for resource gaps 3

3

.

Student Cheating and Academic Integrity Concerns

Most students, 59%, say student cheating with AI happens at least somewhat often at their school, with a third reporting it happens extremely or very often 4

4

. Among teens who use AI for school, three-quarters say cheating is a regular occurrence. However, 15% remain unsure what counts as academic dishonesty, highlighting confusion around acceptable use.

Matthew Kirschenbaum, who teaches English at the University of Virginia and serves on the Modern Language Association's Task Force on AI Research and Teaching, warned that agentic AIs represent a fundamental challenge. "Einstein is symptomatic. I doubt we'll be talking about Einstein, as such, in a year. But it's symptomatic of what's about to descend on higher ed and secondary ed as well" 2

2

. The MLA's statement from October 2025 flagged concerns about agentic browsers that can navigate EdTech platforms and complete assignments without any student involvement.

Educators Debate the Future of Learning

The arrival of tools like Einstein forces educators to confront whether AI represents assistance or substitution. Nicholas DiMaggio, a PhD student at The University of Chicago Booth School of Business, suggested the change may ultimately benefit education. "I think the Canvas method of teaching already has a proclivity for cheating. This change, I think, will ultimately be good because it will force educators to redesign classes to not rely on virtual assignments" 1

1

.

Kirschenbaum argues that agentic AIs are the inevitable conclusion of viewing higher education as a transactive process focused on credentials rather than learning. "What we're finding is that if forms of education can be transacted then we've just about arrived at the point where autonomous software AI agents are capable of performing the transaction on your behalf" 2

2

. He reports that colleagues who have retreated from devices entirely in the classroom find students are almost universally grateful and appreciate the reasoning.

Paliwal frames his creation as liberation from academic labor. "I think we really need to question what learning even is and whether traditional educational institutions are actually helping or harming us," he said 2

2

. Yet this perspective raises questions about critical thinking development and whether automating education undermines the purpose of learning itself. Schools now face decisions about whether to ban such tools outright, integrate them under strict guidelines, or fundamentally rethink how learning is measured in the age of AI.