Poe, owned by the question-and-answer site Quora, and a. AI chatbot platform powered by Andreessen Horowitz invested $75 millionprovides users with downloadable HTML files of articles published by paid journalistic publications.
Invoking the service’s Assistant bot with the URL of this WIRED story about an AI-powered search service, such as Confusion plagiarizing one of our stories, provides a detailed, 235-word summary and 1MB of information. file contains an HTML image of the entire article that users can download from the Poe servers directly from the chatbot.
WIRED was similarly able to retrieve articles from paid sites including The New York Times, Bloomberg Businessweek, The Atlantic, Forbes, Defector and 404 Media in downloadable format simply by entering URLs into the Assistant bot’s interface. It appears to be the latest example of the AI industry’s brazen approach to intellectual property, which is rapidly disrupting existing business models in fields such as journalism and music.
“This is a significant copyright issue,” James Grimmelmann, a professor of digital and information law at Cornell University, said in an email. “Because they made a copy on their server, that’s copyright infringement in the first place.” (Quora disputes this, comparing Po to a cloud storage service.)
When asked to summarize the content of a test website run by my colleague Dhruv Mehrotra, the bot did not return a summary but returned an HTML file. According to the website’s server logs, a server identifying itself as “Quora Bot” visited the site immediately after the Assistant bot was prompted to summarize the site. It did not attempt to access the site’s robots.txt page, suggesting that Po and Quora are ignoring the Robots Exception Protocol, a widely accepted but not legally binding web standard.
A prominent media executive, speaking on condition of anonymity to discuss candidly the legally sensitive matter his company is actively investigating, says the publication also observed servers identifying their site as Quora bots immediately after instructing Poe’s chatbot about specific articles; these suggestions, he says, provided most or all of the text of these articles.
“Poe is a platform that allows users to ask questions and engage in back-to-back dialogue with various AI-powered bots provided by third parties,” Quora spokeswoman Autumn Besselman wrote in an email. “We don’t have or train our own AI models. Poe has a feature that allows the user to show the content of a URL to the bot, but the bot will only see the content served by the domain. We would be happy to contact your technical team to ensure that your paid content is not served to people using Poe.”
“File attachments on Poe are created at the behest of users and operate in the same way as cloud storage services, ‘read later’ services and ‘web scissors’ products, all of which are subject to copyright law,” Besselman wrote in his response. to an email asking additional questions. Andreessen Horowitz did not respond to a request for comment.