Elon Musk's social media company X says it will block its AI chatbot Grok from creating explicit images of real people after governments around the world launched investigations into the feature.
MICHEL MARTIN, HOST:
Governments around the world have launched investigations into the social media platform X since it started allowing users to make and publish sexualized images of women and children without their permission. Two governments have even banned the company's AI chatbot, Grok, as a result of that feature. Now, NPR's Huo Jingnan reports that X is introducing new safety updates.
HUO JINGNAN:
After weeks of mostly ignoring and mocking critics, X now says it will bar users from tagging the Grok chatbot and making sexually explicit public images of real people. Previously, users could ask Grok to edit images and put people into revealing clothing like bikinis. Those images are then published for anyone to see on X. X's owner, Elon Musk, challenged users to break Grok's image moderation, which several quickly did. He blamed lapses on, quote, "adversarial hacking" and said any problems would be fixed immediately. Riana Pfefferkorn is a policy fellow at Stanford University.
RIANA PFEFFERKORN:
Clever and motivated users will try to find ways to fool the list that you've already come up with of words and phrases and prompts that shouldn't be complied with, and come up with other proxies for those same concepts in order to get around the safeguards that you're trying to build in.