OpenAI Disables Bing After Users Bypass Paid Content

OpenAI disabled ChatGPT’s Browse, a search feature based on Bing, after it allowed users to bypass paywalled content, Cointelegraph wrote.

On July 4, the creators of ChatGPT tweeted about the temporary suspension so they could solve the issue and make it up to content owners. OpenAI explained that ChatGPT's ‘Browse’ beta “occasionally displayed content” in unwanted ways, such as fulfilling requests for a URL's full text. They will suspend Browse while the issue is fixed.

Reddit users flagged the problem first

Browse is currently in its beta version and ChatGPT Plus users have access to it. At the end of June, a Reddit user posted a screenshot of a Browse session, where the bot gave them access to the text of an article published in The Atlantic, available to paying customers only. ChatGPT provided the article in full.

Bot scrapes text behind paywalls

The post received hundreds of comments. Apparently, ChatGPT works similarly to online paywall removers, which access content cached by Google. This content has no paywall for SEO purposes.

A Reddit user opined ChatGPT might be capable of ignoring paywall code used to hide the top of the content until a user has either logged in or signed up.

Other issues with data scraping

Twitter owner Elon Musk imposed new limits on the number of tweets users can access per day because of data scraping. What’s more, this is not the first time OpenAI has had issues with scraping. A class-actionlawsuitwas filed in California against the company over scraping private user information from the internet.

The lawsuit, filed on June 28, alleges that OpenAI trained ChatGPT using data from millions of Wikipedia articles, blogs, and social media comments without users’ consent, thus violating the privacy of millions.

The risks

As Bankless Times wrote, over 100,000 ChatGPT login credentials were leaked and traded on the dark web from June 2022 to May 2023 according to Group IB, a Singaporean cybersecurity firm.