AI Potholes
Explore the road ahead to learn how to avoid common risks related to accuracy, security, learning, and ethics so you can use your travel companion, Artificial Intelegence, responsibly.
Security Potholes
Learning Potholes
Ethics Potholes
Accuracy Potholes
Learning Potholes
Relying too heavily on AI tools to complete challenging intellectual and creative tasks can hinder your own learning and lead you to produce inferior work.
AI tools can enable cognitive offloading, or completing tasks with little effort while conserving your thinking power for higher priority tasks. However, that reduced effort can lead to cognitive debt, or long-term negative consequences to memory, critical thinking, and the development of expertise. If you aren't engaging your brain in challenging tasks, your capacity for independent problem solving may fail to grow or even begin to diminish.
Cognitive Debt
AI tools often produce polished, well-formatted, coherent answers that feel correct, making it difficult to detect errors or omissions. An AI user may mistake an apparently high-quality AI result for their own competence, giving them a sense of mastery and confidence that masks what may be shallow understanding. Without human intelligence and judgement that comes from genuine learning, you might submit documents with critical errors to your teacher or your employer and be held accountable for shoddy work.
The Illusion of Competence
Because generative AI produces results based on pattern matching and probability, its results are inherently derivative. That means it is a remix of what has come before, not an original creation. When millions of people use the same tools to create text and images with little human intervention, the results that are shared can be repetitive and generic. Humans can draw on lived experience and emotion to enhance their work in a way that AI tools cannot. Just as critical thinking skills need to be developed through challenging practice, your creativity also needs to be nurtured and exercised. Over-reliance on AI can hinder that progress.
LAck of Originality
Ethics Potholes
The ethical implications of AI use are complex and present new challenges that we must use human judgement to address. You are ultimately responsible for the choices you make and their unintended consequences as you engage this emerging technology.
ECU's Academic Integrity policy identifies "using generative AI tools (such as chatbots) to produce any content submitted for an assignment without the instructor’s permission" as an act of academic dishonesty and a violation of ECU's core values. Unauthorized use of AI tools can make it difficult for instructors to assess what you have actually learned in the course. As a student, you must be honest about the work you complete, follow instructions and course policies, acknowledge the work of others, and disclose assistance you have received from AI tools. Students are responsible for knowing and following ECU's Academic Integrity Policy.
Academic Integrity
Academic Integrity Policy
Generative AI has created legal ambiguity over intellectual property when it can be difficult to determine what was created by a human or by a machine. Artists, authors, and content creators are increasingly suing AI companies for using their work to train AI models without permission. The results of this unauthorized training is that people can easily create images in the style of an artist without giving that artist credit or compensation. It can also cause a student to unintentionally plagiarize the work of another person if the AI tool generates a response very similar to copyrighted material. Failure to protect intellectual property means that people have less incentive to create new things, from novels to paintings to new inventions that can improve our lives.
Intelectual Property
The power demands of AI data centers are rapidly increasing, as is the AI environmental footprint. Data centers require large volumes of water for cooling, and server hardware depends on the mining of rare earth minerals. AI infrastructure relies on grids powered by fossil fuels, contributing to carbon emissions. As the AI server industry continues to grow, we must remain aware of its environmental impact and expand energy usage responsibly.
Environmental Impact
It can be easy to assume that because AI tools are machines, they are inherently neutral. In fact, AI bots have replicated the biases found in their training materials. If large language models are trained on data that is not representative or that contains stereotypes or historical bias, the AI output will also be biased. For example, if an AI tool used to screen resumes is trained on historical data from a male-dominated field, it may perpetuate that bias by downgrading resumes with language it associates with women. Human intervention and evaluation is required to ensure that the results that come from AI tools don't reflect bias or stereotypes.
Bias
Security Potholes
The rapid adoption of AI has introduced security risks connected to student and faculty data and vulnerability to malicious actors who use AI to deceive others.
Many AI platforms, especially widely-available free versions, use the information that users input into the tool to train their Large Language Models (LLM). When you enter an assignment prompt or your own essay into an AI tool, it often becomes part of the AI's knowledge base and can potentially be exposed to others. For this reason, many companies have policies against the use of AI in their workplace because they want to protect their proprietary information.
Data Usage for Training
Entering personally identifiable information (PII) like names, birthdates, or grades can be especially dangerous and can violate privacy laws like FERPA. Many people overlook the risk associated with AI assistants that are designed to listen to a meeting and create summary notes. In those cases, the tool can create a written document that is then emailed to other attendees, documenting and sharing information that was intended to be private.
Personally Identifiable Information (PII)
Generative AI tools have made widespread phishing, deepfakes, and malware generation faster and easier for attackers who can create highly personalized, convincing emails, videos, and phone calls that lure people into sharing sensitive information. Protecting yourself from digital scammers requires continuous learning and consistent attention to warning signs.
Cyberattacks
Accuracy Potholes
Generative AI tools can produce inaccurate results, so human engagement is essential at every step of the process. Passive acceptance of AI-generated content can lead to misunderstanding, misinformation, and misrepresentation. Ultimately, you are responsible for the information included in your work, even if the AI tools have led you astray.
AI Tools like ChatGPT, Copilot, and Google Gemini can produce confident, authoritative answers that are entirely inaccurate or fabricated. Using a chatbot as though it is a search engine, or even relying on the "AI Overview" at the top of your Google results, can lead you to believe and then share inaccurate information.
Hallucinations
Chatbots frequently hallucinate citations, so if you prompt a bot to cite its sources, chances are they don't exist. Using a chatbot in place of a reliable database, like those found on the Linscheid Library website, will lead to a dead end in your research, or even to a failing grade on your research paper!
Fake Citations
Linscheid Library
Library Databases
Large Language Models (LLM) are relying on information that they been trained on, which includes a great deal of information that is no longer accurate. They do not show you the dates of the sources they are drawing from or credentials of the authors of those sources, so you must do some additional research to confirm that the information presented is current.
Outdated Information
AI tools complete tasks through pattern recognition rather than true understanding. They may present unfounded or oversimplified conclusions with the appearance of confidence, and they often do not handle uncertainty or nuance well. These tools create the illusion of intelligence, but they lack comprehension.
Weak Reasoning
AI Potholes v2
Skye Norman
Created on May 14, 2026
Start designing with a free template
Discover more than 1500 professional designs like these:
View
Geographical Challenge: Drag to the map
View
Explainer Video: AI for Companies
View
Explainer Video: Keys to Effective Communication
View
SWOT Challenge: Classify Key Factors
View
Big Data: The Data That Drives the World
View
Mind Map: The 4 Pillars of Success
View
Momentum: Employee Introduction Presentation
Explore all templates
Transcript
AI Potholes
Explore the road ahead to learn how to avoid common risks related to accuracy, security, learning, and ethics so you can use your travel companion, Artificial Intelegence, responsibly.
Security Potholes
Learning Potholes
Ethics Potholes
Accuracy Potholes
Learning Potholes
Relying too heavily on AI tools to complete challenging intellectual and creative tasks can hinder your own learning and lead you to produce inferior work.
AI tools can enable cognitive offloading, or completing tasks with little effort while conserving your thinking power for higher priority tasks. However, that reduced effort can lead to cognitive debt, or long-term negative consequences to memory, critical thinking, and the development of expertise. If you aren't engaging your brain in challenging tasks, your capacity for independent problem solving may fail to grow or even begin to diminish.
Cognitive Debt
AI tools often produce polished, well-formatted, coherent answers that feel correct, making it difficult to detect errors or omissions. An AI user may mistake an apparently high-quality AI result for their own competence, giving them a sense of mastery and confidence that masks what may be shallow understanding. Without human intelligence and judgement that comes from genuine learning, you might submit documents with critical errors to your teacher or your employer and be held accountable for shoddy work.
The Illusion of Competence
Because generative AI produces results based on pattern matching and probability, its results are inherently derivative. That means it is a remix of what has come before, not an original creation. When millions of people use the same tools to create text and images with little human intervention, the results that are shared can be repetitive and generic. Humans can draw on lived experience and emotion to enhance their work in a way that AI tools cannot. Just as critical thinking skills need to be developed through challenging practice, your creativity also needs to be nurtured and exercised. Over-reliance on AI can hinder that progress.
LAck of Originality
Ethics Potholes
The ethical implications of AI use are complex and present new challenges that we must use human judgement to address. You are ultimately responsible for the choices you make and their unintended consequences as you engage this emerging technology.
ECU's Academic Integrity policy identifies "using generative AI tools (such as chatbots) to produce any content submitted for an assignment without the instructor’s permission" as an act of academic dishonesty and a violation of ECU's core values. Unauthorized use of AI tools can make it difficult for instructors to assess what you have actually learned in the course. As a student, you must be honest about the work you complete, follow instructions and course policies, acknowledge the work of others, and disclose assistance you have received from AI tools. Students are responsible for knowing and following ECU's Academic Integrity Policy.
Academic Integrity
Academic Integrity Policy
Generative AI has created legal ambiguity over intellectual property when it can be difficult to determine what was created by a human or by a machine. Artists, authors, and content creators are increasingly suing AI companies for using their work to train AI models without permission. The results of this unauthorized training is that people can easily create images in the style of an artist without giving that artist credit or compensation. It can also cause a student to unintentionally plagiarize the work of another person if the AI tool generates a response very similar to copyrighted material. Failure to protect intellectual property means that people have less incentive to create new things, from novels to paintings to new inventions that can improve our lives.
Intelectual Property
The power demands of AI data centers are rapidly increasing, as is the AI environmental footprint. Data centers require large volumes of water for cooling, and server hardware depends on the mining of rare earth minerals. AI infrastructure relies on grids powered by fossil fuels, contributing to carbon emissions. As the AI server industry continues to grow, we must remain aware of its environmental impact and expand energy usage responsibly.
Environmental Impact
It can be easy to assume that because AI tools are machines, they are inherently neutral. In fact, AI bots have replicated the biases found in their training materials. If large language models are trained on data that is not representative or that contains stereotypes or historical bias, the AI output will also be biased. For example, if an AI tool used to screen resumes is trained on historical data from a male-dominated field, it may perpetuate that bias by downgrading resumes with language it associates with women. Human intervention and evaluation is required to ensure that the results that come from AI tools don't reflect bias or stereotypes.
Bias
Security Potholes
The rapid adoption of AI has introduced security risks connected to student and faculty data and vulnerability to malicious actors who use AI to deceive others.
Many AI platforms, especially widely-available free versions, use the information that users input into the tool to train their Large Language Models (LLM). When you enter an assignment prompt or your own essay into an AI tool, it often becomes part of the AI's knowledge base and can potentially be exposed to others. For this reason, many companies have policies against the use of AI in their workplace because they want to protect their proprietary information.
Data Usage for Training
Entering personally identifiable information (PII) like names, birthdates, or grades can be especially dangerous and can violate privacy laws like FERPA. Many people overlook the risk associated with AI assistants that are designed to listen to a meeting and create summary notes. In those cases, the tool can create a written document that is then emailed to other attendees, documenting and sharing information that was intended to be private.
Personally Identifiable Information (PII)
Generative AI tools have made widespread phishing, deepfakes, and malware generation faster and easier for attackers who can create highly personalized, convincing emails, videos, and phone calls that lure people into sharing sensitive information. Protecting yourself from digital scammers requires continuous learning and consistent attention to warning signs.
Cyberattacks
Accuracy Potholes
Generative AI tools can produce inaccurate results, so human engagement is essential at every step of the process. Passive acceptance of AI-generated content can lead to misunderstanding, misinformation, and misrepresentation. Ultimately, you are responsible for the information included in your work, even if the AI tools have led you astray.
AI Tools like ChatGPT, Copilot, and Google Gemini can produce confident, authoritative answers that are entirely inaccurate or fabricated. Using a chatbot as though it is a search engine, or even relying on the "AI Overview" at the top of your Google results, can lead you to believe and then share inaccurate information.
Hallucinations
Chatbots frequently hallucinate citations, so if you prompt a bot to cite its sources, chances are they don't exist. Using a chatbot in place of a reliable database, like those found on the Linscheid Library website, will lead to a dead end in your research, or even to a failing grade on your research paper!
Fake Citations
Linscheid Library
Library Databases
Large Language Models (LLM) are relying on information that they been trained on, which includes a great deal of information that is no longer accurate. They do not show you the dates of the sources they are drawing from or credentials of the authors of those sources, so you must do some additional research to confirm that the information presented is current.
Outdated Information
AI tools complete tasks through pattern recognition rather than true understanding. They may present unfounded or oversimplified conclusions with the appearance of confidence, and they often do not handle uncertainty or nuance well. These tools create the illusion of intelligence, but they lack comprehension.
Weak Reasoning