Atlas, the AI-powered web browser built with ChatGPT at its core, is designed to do more than just let people surf the internet. It can read, summarise, and even complete online tasks like booking appointments or finding hotels on behalf of the user. Well, a new investigation suggests that Atlas browser may be steering clear of websites owned by companies that are suing OpenAI. According to an investigation by Aisvarya Chandrasekar and Klaudia Jazwinska of the Columbia Journalism Review (via Gizmodo), when Atlas operates in its agent mode, it carefully avoids certain sources.
Survey
✅ Thank you for completing the survey!
Normally, web crawlers follow the rules of the internet. If a website tells them not to enter certain pages, they simply don’t. But Atlas’s agent mode doesn’t always follow those limits, according to the report. It can browse pages that block automated access.
However, Chandrasekar and Jazwinska discovered something interesting when they asked Atlas to summarise articles from PCMag and The New York Times, both of which are involved in copyright lawsuits against OpenAI. Instead of directly accessing those sites, Atlas used complicated ways to get the information indirectly.
For the PCMag article, Atlas searched social media posts and other websites quoting or discussing the story. For the New York Times article, it built a summary using information from other outlets like The Guardian, The Washington Post, Reuters, and The Associated Press- all of which have partnerships or agreements with OpenAI, except Reuters.