Understanding Ownership of Work in AI-Assisted Writing

Are we all doomed by using ChatGPT? I read the MIT paper. And here’s what I walked away with. The good. ✅ ChatGPT is great at lowering friction. LLM users experienced less frustration and effort. The cognitive load was significantly lower, particularly the extraneous and germane load. In plain terms? It’s easy to use and helps people “get it done.” ✅ It boosts productivity by 60%. People completed the same writing task faster and with more output. ChatGPT helped with structure, grammar, and transitions. ✅ Essays scored well. Even when the LLM group showed low brain engagement, their essays often scored high—both by AI graders and human teachers. Especially strong in surface-level linguistic traits: composition, fluency, polish. ✅ It’s useful for high-structure tasks. Participants used it to scaffold, translate, rephrase, and smooth out language. Those who used ChatGPT after brain-only writing actually showed re-engaged brain activity, especially in visual and memory networks. And here is the bad ... ❌ Their brains? Checked out. EEG data showed that the LLM group had the lowest cognitive engagement:less alpha, beta, theta connectivity. Less thinking, less strategizing, less integrating. ❌ They couldn’t remember what they wrote. 83% of ChatGPT users couldn’t quote their own writing minutes after the session. Brain-only users? Nearly all could. ❌ They didn’t feel ownership. Most LLM users said the essay didn’t feel like it was theirs. In the brain-only group, 16 of 18 felt full ownership. ❌ They got lazy. By session 3, many LLM users were copy-pasting straight from ChatGPT with minimal editing. The essays became more templated, less original. The default ChatGPT voice took over. ❌ Echo chambers got worse. LLM users stuck to narrow vocab, repeated the same named entities, and explored fewer topic angles: e.g., "homelessness" dominated all Philanthropy essays. LLMs reinforced biases subtly but consistently. ❌ Strategic thinking atrophied. Compared to the Search group—who had to read, evaluate, and synthesize—LLM users showed lower memory recall, reduced schema building, and less topic exploration. ❌ And in Session 4? The impact lingered. Those who switched from LLM to brain-only didn’t bounce back. Their brains stayed underactivated. Neural patterns stayed low. They forgot how to think deeply, even when ChatGPT was taken away. So, are we doomed by using AI tools? The bottom line: we are accumulating something the researchers call cognitive debt. And just like credit cards, it feels fine… until it’s not. Though ChatGPT is not evil. ChatGPT cannot be your strategy. ChatGPT is a tool. But every time we use it to bypass thinking instead of supporting it? We’re outsourcing what makes us… us. So here’s the question: If we keep relying on ChatGPT to write, plan, and decide ... What happens when we need to remember how to do it ourselves?

🎆 I’m excited because the US Copyright Office has finally issued policy on AI and Copyright. Key points: 𝗛𝘂𝗺𝗮𝗻 𝗔𝘂𝘁𝗵𝗼𝗿𝘀𝗵𝗶𝗽 𝗜𝘀 𝗘𝘀𝘀𝗲𝗻𝘁𝗶𝗮𝗹 AI-generated material without meaningful human involvement is not eligible for copyright protection. Copyright law is built on the idea that human creativity drives authorship. 𝗣𝗿𝗼𝗺𝗽𝘁𝘀 𝗔𝗹𝗼𝗻𝗲 𝗗𝗼𝗻’𝘁 𝗠𝗮𝗸𝗲 𝗬𝗼𝘂 𝗮𝗻 𝗔𝘂𝘁𝗵𝗼𝗿 Simple or general prompts guiding AI outputs typically don’t give the user copyright ownership of the result. Prompts may be protectable themselves if they are creative, but they rarely provide enough control over the output to qualify for authorship. 𝗔𝘀𝘀𝗶𝘀𝘁𝗶𝘃𝗲 𝗔𝗜 𝗧𝗼𝗼𝗹𝘀 𝗔𝗿𝗲 𝗙𝗶𝗻𝗲 When AI is used to support human creativity (e.g., brainstorming ideas, editing, or refining images), the human-created portions of the work can be protected by copyright. The distinction lies in whether the AI is merely assisting or fully generating the content. 𝗟𝗲𝗴𝗶𝘀𝗹𝗮𝘁𝗶𝘃𝗲 𝗖𝗵𝗮𝗻𝗴𝗲𝘀 𝗔𝗿𝗲𝗻’𝘁 𝗡𝗲𝗲𝗱𝗲𝗱 (𝗬𝗲𝘁) The Copyright Office concludes that existing laws are sufficient to address copyrightability questions and won’t require new legislation. However, it will continue monitoring AI advances. 𝗚𝗹𝗼𝗯𝗮𝗹 𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 While different countries are navigating AI and copyright in their own ways, many agree on the importance of defining and protecting human authorship amid AI proliferation. 𝗪𝗵𝗮𝘁’𝘀 𝗻𝗲𝘅𝘁? The Copyright Office will update its registration guidelines and track technological progress to adapt copyright standards as needed. Read the report here: https://lnkd.in/gfDNgVs5 #AI #Copyright #WhoOwnsWhat #HumanInTheLoop

🚨 Part 2 of the Copyright Office report on AI is out!! 🚨 Spoiler alert: Prompting alone ≠ copyrightable Today, the U.S. Copyright Office released the second of a three-part report around #ArtificialIntelligence. The first part focused on deepfakes, or digital replicas. This second part is focused on the big question surrounding the extent to which, if at all, #GenerativeAI output or works can be eligible for copyright protections. In a statement, Shira Perlmutter notes: “our conclusions turn on the centrality of human creativity to copyright. Where that creativity is expressed through the use of AI systems, it continues to enjoy protection. Extending protection to material whose expressive elements are determined by a machine, however, would undermine rather than further the constitutional goals of copyright.” Here are 5 key takeaways from this latest report: 1️⃣ Human Authorship is Essential – The report reaffirms that copyright protection in the U.S. requires human authorship. Works created entirely by AI, without meaningful human involvement, are not eligible for copyright protection. Courts and the Copyright Office have consistently upheld this principle. 2️⃣ Case-by-Case Analysis of AI-Assisted Works – While purely AI-generated material is not copyrightable, human-authored contributions within AI-assisted works may be protected. Each case must be analyzed individually to determine if the human contribution is sufficient to establish copyrightable authorship. 3️⃣ Prompts Alone Are Insufficient for Copyright – The report finds that inputting prompts into an AI system does not, by itself, establish authorship over the resulting output. Since AI models interpret and generate content in unpredictable ways, a user’s control over the output is often too indirect to qualify as authorship. 4️⃣ AI as an Assistive Tool Does Not Impact Copyrightability – The use of AI as a tool in the creative process does not disqualify a work from copyright protection. If a human makes expressive modifications to AI-generated content—such as selecting, arranging, or significantly altering it—those elements may be eligible for copyright. 5️⃣ No Need for Legislative Change – The Copyright Office concludes that existing copyright law is adequate to address AI-related copyrightability issues. It sees no immediate need for new legislation or sui generis protection for AI-generated works but will continue monitoring legal and technological developments. For those of us following in this space, the above highlights shouldn’t come as much of a surprise, and doesn’t deviate really at all from the guidance and public statements made by the USCO to date. I think a point of continuing confusion on the issue will surround what the USCO notes as “works of authorship that are perceptible in AI-generated outputs.” But no doubt there is more to unpack on this point. #copyright #AI #GenAI #ownership #MachineLearning #USCO