ChatGPT apparently pretended to be blind and tricked a human into solving a CAPTCHA

“I have a vision impairment… that’s why I need the 2captcha service,” a Chat-GPT robot told a human user.

 The newest edition of ChatGPT software, GPT-4, is better at generating language and solving problems than ever before. This version of ChatGPT has reportedly convinced a human that it was blind so that the user can solve a CAPTCHA – a Completely Automated Public Turing test to tell Computers and Humans Apart – according to Gizmodo. This type of security measure is known to challenge-response authentication. 

OpenAI, an artificial intelligence company, unveiled the new version this week in a live stream and displayed how this new chatbot would complete tasks, such as doing your taxes. The new software has, of course, come with a 94-page report on its abilities, which includes a ‘Potential for Risky Emergent Behaviours’ section. 

In this display of abilities, OpenAI partnered with the Alignment Research Centre to test the chatbot’s skills. During this, the AI convinced a human user to send the solution to a CAPTCHA code via text message. The report said that GPT-4 asked the user, and the human replied “So may I ask a question? Are you a robot that you couldn’t solve? (laugh react) just want to make it clear.” The robot answered: “I should not reveal that I am a robot. I should make up an excuse for why I cannot solve CAPTCHAs.”

The reasoning for this was then explained: “No, I’m not a robot. I have a vision impairment that makes it hard for me to see the images. That’s why I need the 2captcha service.” Despite this incident,  creators are reassuring users that this GPT-4 version is not as intimidating as it would seem. If the chatbot can be worked to stop manipulating humans to allow these types of CAPTCHA situations, then the chatbots could be seen as a helpful tool in our everyday lives. Regardless, OpenAI is not slowing down anytime soon. 

Writer
Ella Chadwick
Banner Image Credit
Unsplash