Regarding AI: "Khan Academy's Seemingly Paradoxical Position on AI Content" and the "Proper Approach to Flagging Generated Content on Khan Academy"

Hello community and guardians! I will attempt to write the following un-opinionated. If you wish to question parts of my post then feel free to ask me and I will gladly supply you with scientific papers on the subject in question.

With the introduction of Algorithmic Intelligence (Inaccurately called Artificiality Intelligence by the unintelligent), the online population has been sold into the world of Large Language Models by OpenAI, Google, etc. Making it a online "trend". While this isn't entirely a bad thing, it also is creating certain issues here on the Khan Academy Computer Science platform. While trying to understand Khan Academy's view on AI I have been trying to wrap my mind around the assumptions that the KA Code of Conduct have.

Regarding: "AI for Learning" and "Legal Ownership of Generated Content"

The first issue I am seeing is Khan Academy users generating content with LLMs and presenting it as their own code, without credit or mention of a LLM being involved. Since KA is a learning platform, this is a definite issue. Why allow AI generated content on a site that promotes learning? It is scientifically proven that AI is slowing learning down instead of accelerating it. [(Gerlich, 2025), (Malik et al., 2025), (Bai, Liu, & Su, 2023), (Mahlangu, 2025)]. I would like to emphasize that these papers state that LLMs are slowing down learning by a almost complete margin when being used as a "crutch" and that it is rarely leveraged even by adults in a responsible, learning orientated, way. So, granted that LLMs can have positive effects on learning (like personalization) Does the KA Community seriously think that kids can be trusted to responsibly use AI? I doubt it. I have seen first hand the amount of problems LLM usage in a classroom environment has created (LLM usage in teaching is a huge issue for me, because AI takes all the fun out of learning and makes it feel impersonal by making "educational content" boilerplate). The legitimacy of AI for education aside, the question on generated code arises. Who actually owns the content generated? Can the user who generated it actually have legal claim to the generated content as theirs while it is widely known that AI is "fundamentally algorithmic, relying on patterns from vast datasets of existing code (like GitHub), leading to significant debates and research on its originality, copyright, and ethical legitimacy?" Tools like GitHub Copilot (I'll bring up a external LLM) are trained on public code repositories, meaning much of the generated output is derived from or closely resembles existing human-written code, often with minor modifications. You can try it out for yourself and see the list of "dependencies/sources" it includes at the bottom of a generated code snippet. So why does the Terms of Use state the following?

6.1. User Content and Ownership.

The Services include features that allow Users to [...]

You acknowledge that you are solely responsible for your User Content and the consequences of posting, creating, or publishing such User Content. For clarity, AI Content is User Content. See Section 10 for additional terms relating to use of AI-Enabled features to generate User Content.

10.3.3. User-Generated Content. [...] Input and Output are collectively “AI Content”. AI Content is subject to the terms applicable to User Content set forth in Section 6 (User Content; License Grant). For clarity, AI Content constitutes User Content. You acknowledge that due to the nature of machine learning and the technology powering AI-Enabled Features, Output may not be unique, and an AI-Enabled Feature may generate the same or similar output for you or your authorized Users as it provides to other Users. For clarity, Outputs for other Users are not considered your (or your authorized Users’) User Content. You also acknowledge and agree that any Outputs may not be protectable under copyright or other intellectual property, proprietary, or other law. We make no warranties or representations, express or implied, that Outputs are protectable under any law. Without limiting these Terms, you are solely responsible for the Input that you and your authorized Users provide and for the use of Output by you and your authorized Users, including for ensuring that you or they have rights to submit the Input and that use of Input does not infringe upon the intellectual property rights of others or violate any applicable law or these terms.

[...]

It states "AI Content constitutes User Content," (which is the legal equivalent of saying the "User retains all ownership rights" under 6.1. User Content and Ownership) but later says: "Outputs may not be protectable under copyright or other intellectual property, proprietary, or other law," and that "Input [must] not infringe upon the intellectual property rights of others or violate any applicable law or these terms." This is counter intuitive. It is saying that Users can own generated content as being created by them, but adds a disclaimer saying it might not hold up under legal scrutiny. This is a paradox, why tell users they own their prompted generated code when it won't hold up under the current United States Law! The U.S. Copyright Office and courts consistently hold that purely AI-generated code isn't copyrightable due to lack of human authorship, but you can claim copyright over the human-authored parts of code created with AI assistance, provided you disclose the AI use and the human contribution is significant enough to meet traditional authorship standards. Simply prompting an AI isn't enough! You must show substantial human creative input to claim ownership, otherwise, the AI-generated bits are public domain (Digital Millennium Copyright Act (DMCA)). Public domain means no copyright protection, allowing free use without permission, though original authors/sources for other public domain works still deserve moral recognition, even if not legally required. What does this mean for Khan Academy Users claiming generated code as their own? As made clear by the KA Terms of Use the act of putting code into your program and hitting save is the equivalent of claiming it as your own under the MIT License. Therefore, KA Users should not legally be allowed to use generated code without proper notation regarding the origin of the code and the sources the LLM parsed it from. Correct?

Regarding: "Proper Approach to Flagging Generated Content on Khan Academy" and "Continuing Forward"

If so, there should be an active way for the community to participate in taking down generated content that has claimed authorship. And, while I have been told that flagging the program and marking it as plagiarism will be adequate, it seems that generated programs have largely been given the blind eye (perhaps due to the presence of Khanmigo and the obvious paradox in the terms of use), so I have been commenting on programs and asking the user why they generated it (sometimes seemingly bluntly, which I apologize for) and as a subtle way for other people to recognize the fact. It seems, however, that this practice of confrontation is highly discouraged by the KA guardian leadership, something of which I was not aware. The issue is, that KA doesn't really have a clear stance on "AI generated content". Neither are they making their discourage of "AI generated content" obvious. It would seem warranted to included a check box to mark a program as "generated", to have a clear, unobfuscated, terms of use regarding the subject in question, and to have posted a clear notice regarding "AI generated content," particularly in the Computer Science section.

Disclaimer and Conclusion:

I am not, in any way, shape, or form trying to perpetrate against Khan Academy. I love Khan Academy very much, I have two accounts, one I used for years for school and the other with which I program and teach other people programming. I am attempting to ask for clarification on the Khan Academy policy so that I may link this discussion when explaining to other people why they shouldn't use AI generated code and claim it as their own. I am very open to other peoples thoughts and direction in regard to this topic, and civilized pointing out of the flaws in my logic. I am also not speaking of leveraging AI to assist with coding or with debugging, or even learning here on KA via Khanmigo. I believe those are all relevant, responsible, and good uses of AI. I am specifically regarding users stating that AI generated content is their own, the unclear terms of use, the lack of communication, and the legal follow-throughs (such as unspecified content being automatically plagiarism due to it being automatically under the MIT License). I hope nothing I have said was offensive or misleading to anyone, and am sure that there will be a reasonable logic in what has happened the past few month

I hope you all have a wonderful weekend!

Thoughtfully yours, Arrow

William I would be honored if you participated in this discussion

JaniceHolz

January 05, 2026 23:54
Edited

My understanding is that KA does not want ordinary users to confront other users on any issue, but flag the post and let KA guardians handle it. (IMO this is not a great solution when it sometimes takes weeks for KA guardians to review flagged posts. Yes, in July I went to the trouble of tracking posts I flagged the same day they were made, and most of them were still flagged 10-20 days later -- I expect that the flag gets removed when the post gets reviewed, so that seems to be evidence that it can take a long time for posts to get flagged.)

I agree that AI use should always be documented, and that not doing so is plagiarism. If that is not already clear in KA's community standards, that is one of many things they should make clearer.
Under KA's current system, I flag the post/program and sometimes add a comment as to what source was copied or what my evidence is that the post/program was at least partially AI-generated.

~ Janice, KA user

William

January 11, 2026 23:52

Thanks for raising these points and initiating a meaningful discussion about how we can facilitate effective learning outcomes with AI! Earlier last week, I forwarded your message to our Community team, who have already spent significant time evaluating what helpful guidance in this area could look like. Ultimately, creating and enforcing a policy is challenging if we can't reliably identify whether a project was "assisted with AI" versus "generated with AI." With AI becoming much more powerful, it's become an increasingly complex and speculative process to identify these types of projects.

If you encounter a project that is clearly AI-generated, we encourage you to flag the project and privately write comments as needed for Guardian review. Asking a user why they didn't provide an AI citation is considered unhelpful (and hence off-topic) because it doesn't contribute meaningfully to student learning, and it typically involves accusatory language, which is harmful for an educational setting. When a moderator hides a project, we privately offer students additional guidance to encourage meaningful engagement within our community.

I hope that this provides some clarity. We're also learning, so our approach won't necessarily be perfect, but we hope to continue improving how we support students in leveraging AI responsibly.

January 12, 2026 19:36

I get frustrated with how long it takes for flagged content to get reviewed. I haven't been looking at or flagging computer programs, but it has often taken over a week for Questions/Answers/Comments I flagged to get reviewed (even if I flagged them the same day they were posted). It is also frustrating that the flagger is never given any feedback or guidance on whether their flagging is producing any results. The only way I have found to determine that is by painstakingly documenting some of my flags and then checking on them occasionally. It's exhausting and feels thankless.

I think Khan Academy would greatly benefit if more users were providing good Answers and flagging guideline-breaking content. If we had more confidence in the timeliness and usefulness of the review progress, I think most users would be more willing to just flag rather than trying to post something themselves. Would Khan Academy be willing to work with some high-volume helpers (users who make many useful flags or provide many helpful Answers) to make the process more satisfying for helpers?

~ Janice, KA user who spent half of 2025 trying to Answer all genuine new Questions in the 6th grade math course

Arrow

January 16, 2026 03:15
Edited

William Evan Lewis

Thank you for all your helpful responses and reaction to this. I look forward to seeing the plan Evan mentioned was en route. I am very appreciative of the community focused content and support here, you have set an example for the stackexchange as well as other online communities

https://support.khanacademy.org/hc/en-us/articles/202912080-How-can-I-become-a-Guardian-and-what-does-this-entail. Perhaps the best way to make change is to do it yourself :)

Yours,

January 17, 2026 21:55

I already sent in my application months ago (back in July or earlier). They only recruit about once/year, and I have not heard back. (I think I've suggested that they add something to that specific post as to when their last recruitment was, so people who want to apply have some idea as to when they might hear something.)

I have volunteered for Khan Academy in the past (6 years ago with evaluating JavaScript projects and 11 years ago with captioning videos). The people I worked with seemed to appreciate my efforts.

Regarding AI: "Khan Academy's Seemingly Paradoxical Position on AI Content" and the "Proper Approach to Flagging Generated Content on Khan Academy"

Regarding: "AI for Learning" and "Legal Ownership of Generated Content"

Regarding: "Proper Approach to Flagging Generated Content on Khan Academy" and "Continuing Forward"

Disclaimer and Conclusion:

Didn't find what you were looking for?

5 comments

Didn't find what you were looking for?