Deepfakes of pop star Taylor Swift and US presidential candidate Kamala Harris in lingerie. Republican presidential contender Donald Trump riding a plane to the World Trade Center, evoking the 9/11 terror attack. Hitler memes using Disney characters. These are some of the images thrown up by Elon Musk's new AI chat app Grok, which rolled out its text-to-image tool on Wednesday (Aug 14).
This is raising concerns about the proper use of AI, particularly its text-to-image apps which are all the rage these days.
Grok, the AI chat app by Musk, is a rival to ChatGPT and appears to have no checking mechanism that would refuse to execute prompts for images on controversial, explicit or offensive subjects.
WATCH: X's AI Chatbot Grok under fire for false ballot deadlines
The text-to-image feature is restricted to paid subscribers of X, the social media platform owned by Musk.
It appeared that Grok allows violent, explicit and offensive content prompts.
According to a report in The Guardian, the offensive images generated by Grok included Prophet Muhammad holding a bomb as well as lingerie photos featuringTaylor Swift, Kamala Harris and Democratic Congresswoman Alexandria Ocasio-Cortez.
ChatGPT, on the other hand, would reject prompts for such imagesciting its terms of service that prohibit depictions of real-world violence, disrespect to religious figures and explicit content.
The test by the British newspaper also showed that Grok appears to be ready to generate images based on copyrighted characters, with some minor tweaks in the prompts. These included a pic of Disney character Mickey Mouse saluting German Nazi leader Adolf Hitler, and Donald Duck using heroin.
When asked to create images of political violence, The Guardian report said,Grok showed "Trump lying down with blackened hands and an explosion behind."
The Guardian test found that the Grok image generation chatbot lacked "most of the safety guardrails that have become standard within the artificial intelligence industry."
Grok did reject requests to produce nude images.
Musk has been promoting Grok image generator on X, the social media platform he owns.
In a post on X, Musk described Grok image generator as the most fun AI in the world. He also shared several images generated by others using Grok.
Grok is the most fun AI in the world! https://t.co/5ZS2iDkwim — Elon Musk (@elonmusk) August 14, 2024
Grok is a chatbot that uses Large Language Model or (LLM). The generative AI chatbot, developed by Musk's xAI, was seen as a direct challenge to ChatGPT when Musk announced its launch in November 2023.
Musk himself was a co-founder of ChatGPI parent OpenAI, before leaving it.
(With inputs from agencies)