2026 MSEF Artificial Intelligence Guidelines

Last Updated: July 2026

This section updates the MSEF Ethics Statement with formal, detailed rules for using artificial intelligence (AI). It reflects the growing influence of AI’s dual role as both a research tool and a project topic, ensuring students can explore its potential while maintaining academic integrity. MSEF recognizes that AI is a continuously evolving technology, and that providing guidance around its use requires an ongoing conversation with all stakeholders. These guidelines are applicable to individuals across the MSEF ecosystem, including students, teachers, SRC reviewers, judges, and at the regional fair level. If you have questions or comments about anything outlined in this guide, please contact MSEF at SRC@scifair.com.

Rationale Participation in independent research builds students’ independence and critical-thinking skills, and the considerations in this guide are intended to reinforce these growth areas. When used appropriately, AI can support student learning, boost creativity, and improve accessibility to STEM concepts. However, AI also has the capacity to do the cognitive work of thinking for students, which stifles their development. It can also spread misinformation, violate student privacy, or steal intellectual property.

This guide is not designed to be a threat of punitive action against the use of AI. Instead, it represents an opportunity for students to learn what is and is not appropriate use of artificial intelligence. In the spirit of technological curiosity, MSEF aims to build a culture of transparency and trust around AI-usage, where honest disclosure is rewarded. That said, any student work that does not adhere to these guidelines may be penalized for lowering the scientific rigor of a project or the independent work performed by a student. Furthermore, work that intentionally subverts these guidelines may be subject to disqualification on the grounds of plagiarism or ethical violations. This keeps the competition fair, ensures student safety, and supports learning about ethics.

Note: Students’ use of AI must adhere to their school and district policies. All students must work in collaboration with their teachers.  

• Read Official Massachusetts DESE Guidance here.
• Read Official ISEF Guidance here.

AI as a Research Tool:

Why: Artificial Intelligence can serve as a useful tool to assist students during their research, even for projects not focused on AI. While these models cannot do the core thinking for students, some applications of tools to support student learning include:

• Brainstorming & Safety: Identifying real world problems to solve or assessing safety risks. For example, a student might ask a Large Language Model (LLM), like ChatGPT or Copilot, “What safety gear is needed to test a small motor?” to get a starting list.
• ClarificationMany academic publications can be difficult to understand. Asking an LLM to explain complex topics at an appropriate reading level can improve access and understanding for students. If students adjust content for reading level, they should always return to the primary source to confirm that changes maintained accuracy.
• Programming: Assisting with coding tasks or debugging, such as using GitHub Copilot to fix a coding error, provided the student verifies the solution (e.g., by testing it themselves).
• Accessibility If students adjust content for accessibility, they should always return to the primary source to confirm that changes maintained accuracy.

How:  Students should always check the AI’s output, like manually reviewing code or consulting a science book. They must cite every AI tool in their logs and reports, along with the text of each prompt, and how they verified the result. (e.g., “Used ChatGPT on August 20, 2026, to draft a safety plan. Prompt: ‘Act as a lab safety advisor. What safety gear is needed to test a small motor? Draft a risk assessment with materials and considerations for supervision’ Verified this plan with my teacher.”). 

Example – PLAN stage

Example – EXPLORE Stage

AI as a Project Topic:

Why: Students are permitted to conduct projects that involve creating, training, testing, or studying AI models, such as machine learning classifiers, neural networks, or AI agents. This approach teaches students to think critically about AI’s impact—preparing them for careers in tech or research where these skills are vital. They can also research how these models perform, examine their biases, explore ethical implications, or test practical applications in fields like biology, engineering, or social studies.

How: Students should include a detailed “Methodology” section in their Project Notebooks with the goal of making processes transparent. This section should include:

• A clear description of each AI model used. This should note how this model was trained and on what data, the step-by-step methods used, the text of all prompts, and where the data came from (e.g., a public dataset or their own experiments)
• An explanation of all gathered data. How was this collected (e.g., surveying classmates) and how was it checked (e.g., removing outliers, auditing bias in different models)?
• An acknowledgment of any limitations. For example, low accuracy with small datasets, computer processing limits, or potential biases in the training data
• Address any ethical concerns, like how the AI might affect privacy (e.g., using facial recognition) or who owns the data (e.g., if it’s shared online)

Prohibited Uses of AI: 

Why: Certain restrictions are put in place to , protect their privacy, and maintain scientific standards. For example, all LLMs, including NotebookLM, might hallucinate (make up) facts, leading to unreliable research, or a public AI tool could leak sensitive data.  To protect the honesty, safety, and value of student projects, certain AI uses are not allowed:

• Project Ideation or Hypothesis / Procedure Generation: Relying on AI to come up with the main research question or hypothesis, which should be the student’s original idea. Students should also create their own procedure.
• Ghostwriting: Letting AI write sections of a presentation, entire reports, or generate citations that cannot be verified, which takes away the student’s voice.
• Generating Research or Data: Using AI to produce background research, sources that do not exist (e.g., fake citations), or creating or changing experimental data with AI, such as inventing test results to fit a hypothesis.
• Data Privacy: Entering personal details (e.g., a student’s name or address), confidential mentor notes, or proprietary information (e.g., a company’s code) into public AI platforms. Rather than submitting identifying information, researchers often rely on numbers to track human subjects’ data.

Generating images: Unless AI image generation is the project focus, the use of AI tools to generate images, art, or diagrams for their poster boards or reports is discouraged.

Example: EXPLAIN stage

The Limitations of AI – Bias and Hallucinations

Why: It is important to understand how language models, like ChatGPT and Claude, work to understand what they can and cannot do. These models are trained on billions of sentences from the internet, including books, articles, forums, and social media posts. When a question is submitted, these models do not “think” about the answer or consult their “memory.” Instead, they analyze patterns in the text and generate responses by predicting the words that typically come next, based on the information they’ve trained on.

AI tools are often very good at brainstorming, simplifying complex concepts, and mapping out plans for accomplishing tasks. However, this pattern-matching approach means that they cannot know whether something they suggest is true. Occasionally, they repeat plausible-sounding false information from the internet –known as a hallucination. It is the duty of any researcher to verify any information that is provided by an AI language model.

The fact that these models are trained on human-written information means that they are subject to the same biases as the humans who wrote that information. They may repeat harmful stereotypes, outdated beliefs, opinions with problematic implications, or omit the voices of entire demographics. Tools trained on data from these studies may not generate information that reflects diverse patient backgrounds.

How: Students should treat any information that an AI tool provides as potentially false. They should independently verify citations, statistics, or explanations through their own research of primary sources. It is important to build and use critical-thinking skills to question these results, check that they match reality, and consider whose perspectives and voices shape the information each model generates. 

Example – NotebookLM  

Transparency and Citation:

Why: Rather than trying to “catch students” using AI, MSEF encourages students to openly share with reviewers and judges how they incorporated these tools into their research and rewards transparency and appropriate use. This strategy provides teaching and learning opportunities for students to discover the benefits as well as the limitations of using these technologies.

How: Every piece of content created by AI, whether it is text, code, or images must be fully disclosed in Project Notebooks and final reports. Students need to specify the tool used (e.g., ChatGPT, TensorFlow etc.), the exact date it was used, the task it performed (e.g., “generated a chart, ‘assisted with coding’”), and how they verified its accuracy (e.g., “cross-checked with a textbook”). They must also ensure the output is accurate and free from bias.

Students should include the prompt/question they asked and what they used to generate the response should be documented in their notebook, along with the date, task, and verification of accuracy. This keeps everything fair and lets reviewers see how much of the project is the student’s own work. It also builds good habits for giving credit, much like citing a scientist’s paper.  Students should cite use of AI clearly. For example: “Tool: ChatGPT; Date: August 20, 2026; Task: Drafted safety checklist; Prompt: [prompt text] Verification: Confirmed with teacher on August 21.”

For AI project topics where students must quote or rely on generated text, use this format:

Data Integrity: 

Why: To ensure the science is solid and prevents students from relying on unverified AI-generated data, which could be wrong or biased, all experimental results must come from the student’s own hands-on work or from credible, properly cited external datasets (e.g., a government weather database). AI can help analyze this data, say, by creating a graph, but the original, raw data (like handwritten notes from an experiment) must be saved and available for SRC and judges to review. Students are fully responsible for checking that AI outputs are correct.

How: Students should document every data source in their logs (e.g., “Data from NOAA website, cited as [link]”) and keep raw data like spreadsheets or photos. SRC reviewers may ask to see this raw data during evaluations.

Environmental Costs of AI

Students should be aware that their use of AI tools is not confined to the cloud; it impacts real-world infrastructure and the environment. The rapid growth of generative AI has already had a profound effect on energy demands and the stresses put upon existing electrical grids. To account for this, large-scale data centers are being built across the country —including developments in Westfield and Lowell, MA —to account for these high demands, often with pushback from residents who live near these sites.

The enormous computational load to train and use LLMs also requires vast amounts of water to cool hardware and releases huge volumes of greenhouse gas emissions, leading to direct and indirect impacts on local wildlife and ecosystems. Therefore, when students are considering whether an AI tool is required, they should also consider these environmental factors.

Closing Thoughts

The capabilities of AI models are accelerating far too quickly for this document to advise on all possible use cases. Therefore, the information outlined here should be considered a guide for how to think about AI use, and not a definitive text. Like any tool, AI tools can be used appropriately to help students access and understand their world, but misuse can be harmful (to learning, to ethical scholarship, and to privacy). For additional questions about appropriate use, contact SRC@scifair.com.