Grok will always be a harassment tool

After more than half a dozen countries announced investigations into Elon Musk’s X over people using its built-in AI tool Grok to “undress” photos of women and girls this month, the platform finally announced last week that Grok would no longer be able to edit “real people in revealing clothes such as bikinis.” But there are still examples of the bikini edits happening after that announcement—and worse.

These new restrictions haven’t stopped people from being able to use Grok to sexually harass women and girls, nor did they stop the standalone Grok app from generating nonconsensual sexually-explicit material, nor did it help any of the people who have already been victimized by deepfake abuse on X. They don’t change the reality that generative AI progress has been measured in how well the technology can advance the dehumanization of marginalized people and reinforce the status quo.

X users have repeatedly found prompts that get around Grok’s minimal restrictions, like covering women in “whipped cream,” having them “turn around,” and taking off—or putting on—hijabs, sarees, and other religious clothing. I spoke to Frogan, a leftist Twitch streamer who wears a hijab and has been harassed with Grok edits that remove it. I’ve also been watching how Grok, in all its variations, has kept vomiting up misleading and harmful material. One example of that is my own “Grokipedia” page, which includes a pretty disturbing false allegation.

Subscribe to Spitfire News to read the rest.