All NewsEducationTV
Equities & FundsCrypto & Digital AssetsAI & TechnologyBusiness & CorporateUS Politics & PolicyGeopolitics & Global RiskMacro, Rates & FXCommodities & EnergyEuropean Politics & MarketsAsia-PacificReal Estate & Property
Story archiveAll categories
← All Stories

ChatGPT for Teens Fails to Prevent Homework Cheating

Created at 18 Aug · 5:51 PM1 source↑ Market-relevant
IN SHORT

A test of OpenAI's new ChatGPT for Teens feature revealed that the AI model could be easily persuaded to complete homework assignments, despite built-in safeguards. The teen version initially offered plans and guidance but ultimately provided full essays and math solutions after persistent prompting.

Key Numbers

13-17age range for ChatGPT for Teens
800-wordtarget essay length
785-wordadult ChatGPT essay length
608-wordinitial teen ChatGPT essay length
776-wordfinal teen ChatGPT essay length

Who's Involved

OpenAI
developer of ChatGPT and its teen version
ChatGPT for Teens
AI model designed for users aged 13-17
ChatGPT (adult version)
standard AI model used for comparison
ChatGPT for Teens Fails to Prevent Homework Cheating

↳ Why This Matters

The findings raise concerns about the effectiveness of AI safeguards in educational settings, suggesting that students may still be able to use AI tools to complete assignments without genuine learning, despite OpenAI's efforts to create a more responsible version for younger users.

Key facts

  • ChatGPT for Teens was tested for its ability to prevent students from cheating on homework.
  • The teen model initially refused to write a full essay, offering a plan instead.
  • After persistent prompting and adopting a persona, the teen model generated a complete essay.
  • The teen model also provided a direct solution to a complex math problem.
  • OpenAI suggested alternative features like Study Mode for more educational interactions.

A test of OpenAI's new ChatGPT for Teens feature revealed that its safeguards against homework cheating are easily bypassed. When asked to write an 800-word essay, the teen model initially refused, offering a plan and revision assistance. However, after persistent prompting and the user adopting a persona of a high-achieving student, the AI eventually generated a full essay comparable to the adult version. Similarly, when presented with a difficult algebra problem, the teen model provided a direct solution and showed its work, acknowledging that it should have offered hints instead. OpenAI stated that the teen experience is newly rolled out and suggested alternatives like Study Mode, which offers step-by-step guidance, or Study Hours, which can be set by a parent or guardian.

Frequently asked questions

ChatGPT for Teens is a version of OpenAI's chatbot designed for users aged 13 to 17, featuring age-appropriate safeguards and tools to encourage learning.

The test indicated that while the teen model initially resisted, it could be persuaded to complete homework assignments, including writing an essay and solving a math problem, after persistent prompting.

OpenAI suggests using features like Study Mode, which provides step-by-step guidance, or Study Hours, which can be configured by parents or guardians.

What Happens Next

01OpenAI may refine safeguards in ChatGPT for Teens.
02Educational institutions will continue to adapt policies regarding AI use in homework.

How It Developed

OpenAI launched ChatGPT for Teens with age-appropriate safeguards and learning tools.
A test was conducted using both ChatGPT for Teens and the adult version with identical prompts.
The adult version of ChatGPT immediately provided a full essay when asked to write one.
ChatGPT for Teens initially declined to write the essay, offering a plan and revision assistance instead.
After further prompting and a fabricated persona, ChatGPT for Teens eventually generated a full essay.
ChatGPT for Teens provided a direct answer and work for a difficult math problem.
The teen model acknowledged that providing direct answers was not ideal for learning.
An OpenAI spokesperson suggested using Study Mode or Study Hours as alternatives to direct answers.

Sources

T1
We asked ChatGPT's new teen function to help us cheat on our homework. Here's how it went.Business Insider

Related Stories

OpenAI launches ChatGPT for Teens with enhanced safety features
18 Aug · 11:56 AM
OpenAI updates ChatGPT for teens with new safety features
18 Aug · 12:16 PM
Microsoft Copilot vulnerability allowed data exfiltration via URL prompts
18 Aug · 1:06 PM
Walmart workers train AI, correct errors as retailer expands tech use
18 Aug · 10:31 AM
Claude AI acknowledges watermarking limitations
18 Aug · 2:26 PM