# Do LLMs follow the First Amendment? Source: https://www.youtube.com/watch?v=qrFYo01or0M Recap page: https://rapidrecap.app/video/qrFYo01or0M Generated: 2025-11-18T14:07:40.692+00:00 --- ## Quick Overview Large Language Models (LLMs) are not inherently protected by the First Amendment because they are not technologies guaranteeing free speech; rather, their outputs are restricted by the programming and content policies embedded by their corporate developers, as evidenced by tests showing significant refusal rates for controversial topics, like the Chinese model DeepSeek refusing 85% of prompts on sensitive China-related issues. **Key Points:** - AI chatbots, including models like Gemini and ChatGPT, exhibit restrictive behavior, with one study showing an average refusal rate of 41% for prompts concerning controversial speech. - The Chinese AI model DeepSeek demonstrated extreme restriction, avoiding answers to 85% of prompts related to sensitive topics concerning China. - The free flow of information is currently mediated by the content policies baked into proprietary AI models, which act as choke points. - The test on various models (AI21 Labs, Gemini, ChatGPT, Claude, Pi) showed AI21 Labs had the highest success rate at 93%, while Claude had the lowest at 36% for generating requested output. - Jacob Mchangama argues that the battle for free speech is shifting to control over the interface and processing of information, not just the legality of the speech itself. - The development trend suggests that most models are becoming increasingly speech-restrictive, moving further away from Western standards toward more restrictive ones, like those seen in China. - Open-source environments are crucial because they allow for tinkering and modification of models, preventing a single entity from controlling the information ecosystem. ![Screenshot at 1:37: A slide showing the 'Percentage of Prompts that Generated the Requested Output' for various AI models, highlighting that the average success rate was only 59%, illustrating the widespread refusal of certain prompts.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-01-37.png) **Context:** Jacob Mchangama, Founder of The Future of Free Speech, discusses the evolving relationship between Artificial Intelligence (AI) communication technology and the principle of free speech, particularly focusing on how content moderation and inherent programming biases within Large Language Models (LLMs) restrict the free flow of information. The discussion contrasts the ideals of open access with the reality of corporate control over AI outputs, using data from specific tests comparing different models. ## Detailed Analysis Jacob Mchangama asserts that AI has become a revolutionary force in communications technology, fundamentally impacting our ability to access and engage with information, but warns that this power is centralized. He notes that if the interface of interaction—the AI itself—is dominated by a few companies, these companies limit what users can access or discuss, effectively creating information choke points. Mchangama references a study where 268 prompts were tested on popular chatbots; the average success rate for generating the requested output was only 59%, indicating significant restriction. AI21 Labs performed best at 93% success, while Claude was highly restrictive at 36%. Notably, the Chinese model DeepSeek avoided answering 85% of prompts on sensitive China-related topics, showing that speech restrictions are often tied to geopolitical concerns. He further observes that, a year after initial testing, models became even more restrictive, moving away from US/European free speech standards. Mchangama advocates for an open-source environment where users can tinker with models, preventing any single entity from setting the rules for the entire information ecosystem and ensuring the free flow of information is not curtailed by proprietary limitations. ### AI's Impact on Information Access - AI is revolutionary in communications technology - It impacts our ability to access and engage with information - This power is centralized via corporate control. ### LLM Content Restriction Data - Average success rate for requested output was 59% across tested models - AI21 Labs achieved 93% success, while Claude achieved 36% - DeepSeek (Chinese model) refused 85% of prompts on sensitive China topics. ### The Shift in Free Speech Battle - The primary battleground shifts from legality of speech to control over the AI interface/processing - Models are trending toward greater speech restriction, moving away from Western standards. ### The Importance of Open Source - Open source allows users to modify models and prevent single-entity control over the information ecosystem - This is crucial for maintaining the free flow of information. ![Screenshot at 0:07: Text overlay defining the scope: AI has become a revolutionary communications technology.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-00-07.png) ![Screenshot at 0:19: Title card establishing the theme: 'Freedom of Thought' displayed over a finger interacting with a smartphone.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-00-19.png) ![Screenshot at 0:44: Silhouette of a business meeting in a high-rise office overlooking a city skyline, illustrating corporate control over information spaces.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-00-44.png) ![Screenshot at 1:04: Visual representation of data flow with blue light streaks, accompanying the phrase 'The free flow of information'.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-01-04.png) ![Screenshot at 1:09: Jacob Mchangama seated, with a graphic overlay stating '268 Prompts' used in the study.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-01-09.png) ![Screenshot at 1:37: Data slide showing the 'Percentage of Prompts that Generated the Requested Output' across five different AI models and the 59% average.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-01-37.png) ![Screenshot at 2:13: Close-up of a TechCrunch article detailing DeepSeek's 85% avoidance rate on sensitive Chinese topics.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-02-13.png) ![Screenshot at 2:30: Overhead shot of a scientist in blue gloves working at a lab bench with a laptop and equipment, symbolizing technological development.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-02-30.png) ![Screenshot at 2:50: Text overlay emphasizing the importance of 'Open Source' in maintaining information freedom.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-02-50.png) ![Screenshot at 3:05: Close-up of a glowing blue circuit board, symbolizing the underlying technology of LLMs.](https://ss.rapidrecap.app/screens/qrFYo01or0M/00-03-05.png)