Meat proxy - new term for sloppy AI users
According to Business Insider, the term caused a stir when it appeared in a blog post by German software engineer Niklas Gruhn in early August, to refer to people who directly use, copy, and send output from AI without checking or editing, or even without understanding.
Gruhn said his friends are quoting verbatim responses from Claude, a chatbot developed by American startup Anthropic, and sharing them in chat groups. "This doesn't add any value. I can talk to Claude myself, it's faster and I can control the context. I don't need a meat proxy as an intermediary," Gruhn wrote.
Gruhn's post on technology forum Hacker News attracted more than 700 comments. Some names in the technology field also quickly responded.
Software developer Simon Willison (UK) evaluates meat proxy as "a great new term". Mehul Agarwal, founder and CEO of Koyal AI, also added this term to the technology dictionary website he developed, SF-isms.
"A code merge request is submitted that looks reasonable, large scale, but the person who 'wrote' it can't explain the 40th line of code," said Steve Mayne, CTO of tax platform Yonda, describing the "flesh and blood AI representative" situation on LinkedIn.
Workplace review processes were designed for a time when “coding was expensive and reading was cheap,” he said, but AI reverses this and large volumes of work can now be completed in a matter of seconds.
On the social network Twitter, the phrase meat proxy receives many supportive comments. "Asking them is just like asking Claude. They don't add value but still exist because sometimes Claude works perfectly fine," said Blinkit engineer Ratul Dawar.
"This is the biggest career risk of 2026. I am witnessing this problem firsthand," software developer Cory House shared.
"The line between using AI to increase productivity and being a 'flesh agent' is very thin. Don't be like that," technology founder Nicholas Charriere wrote.
However, according to Mashable, not every answer that begins with "ChatGPT said..." is worthy of condemnation. Some find it helpful to announce AI-generated answers, helping listeners clearly see where the information comes from. Plus, if users have carefully read the feedback, vetted it, and decided the content is worth sharing, citing the chatbot doesn't necessarily cause trouble.
Meanwhile, Gruhn believes that instead of mindlessly copying content, users should read, understand, verify, and then write the content in their own words. "That effort is the added value you can create," he argues.
Meat proxy is just one of many terms that talk about the dark side of artificial intelligence. Previously, "AI garbage" (AI slop) - content created by AI quickly and cheaply in large quantities, but lacking depth, easily causing offense and sometimes misleading information - also received a lot of criticism. Another term, "cognitive surrender", appeared in recent research from the University of Pennsylvania, referring to the situation where users blindly trust advice from AI even when inaccurate, reducing their ability to think critically.
"Humans have a very strong tendency to save energy. In life, we often use strategies that help us get to the point faster without taking the time to carefully study the information, because this will consume a lot of cognitive energy," Johann Chevalere, a cognitive and social psychology expert at the French National Center for Scientific Research (CNRS), told AFP.
Chevalere said that overuse of AI could promote this trend, reducing human thinking capacity. "If there are activities you never do, the brain, which operates on the principle of energy conservation, will not waste time maintaining unused connections," he explains.