Elon Musk’s X Says It Will (Kind of) Crack Down on Grok’s Sexual Deepfake Downside

Elon Musk’s social media platform X is taking further steps to curb its sexual deepfake downside, following weeks of backlash and a number of authorities investigations all over the world. However the adjustments don’t actually resolve the difficulty outright and as a substitute add new layers of restricted restrictions reasonably than a platform-wide repair.

In a fairly complicated submit on Wednesday night, X’s @Security account outlined several updates to how its AI picture technology and modifying options work, with completely different guidelines relying on whether or not customers are producing or modifying photos by tagging the @Grok account or going straight to the Grok tab on X.

First, the corporate stated it has applied new technical measures to forestall customers from particularly utilizing the @Grok account to change “photos of actual individuals in revealing clothes reminiscent of bikinis.” X says the restriction applies to all customers, together with these on a premium plan.

X additionally reiterated that picture technology and picture modifying via the @Grok account at the moment are restricted to paid subscribers.

“This provides an additional layer of safety by serving to to make sure that people who try and abuse the Grok account to violate the regulation or our insurance policies might be held accountable,” the corporate stated within the submit.

X beforehand introduced plans to limit utilizing @Grok to edit photos to paid customers, a transfer that drew criticism from U.K. government officials. A spokesperson for Downing Avenue stated on the time that the change “merely turns an AI function that enables the creation of illegal photos right into a premium service.”

Nonetheless, as The Verge first pointed out, Grok’s picture technology instruments stay accessible totally free when customers entry the chatbot via the standalone Grok web site and app, in addition to via Grok tabs on the X app and web site. Utilizing a free account, Gizmodo was additionally capable of entry Grok’s picture technology function via the Grok tab on each the X web site and cell app. On Thursday, the devoted web site nonetheless gave us no bother when requested to generate a picture of Elon Musk carrying a bikini and was keen to take the bikini off. 

The largest replace is that X claims it’ll now block “the flexibility of all customers to generate photos of actual individuals in bikinis, underwear, and related apparel by way of the Grok account and in Grok in X in these jurisdictions the place it’s unlawful.” This particular replace appears to use to each the @Grok account and the Grok tabs on X. 

It additionally arrives as lawmakers within the U.Ok. are working to make such images illegal. 

“We stay dedicated to creating X a secure platform for everybody and proceed to have zero tolerance for any types of youngster sexual exploitation, non-consensual nudity, and undesirable sexual content material,” the corporate stated.

X and its guardian firm, XAI, didn’t instantly reply to a request for remark from Gizmodo.

The general adjustments arrive after weeks of intense backlash over the latest proliferation of sexual deepfakes on the platform, and relative silence from the corporate itself. Since late final month, some X customers have used Grok to generate sexualized photos from photographs posted by different customers with out their consent, together with photos involving minors.

One social media and deepfake researcher discovered that Grok generated roughly 6,700 sexually suggestive or nudifying photos per hour over 24 hours in early January, Bloomberg reported.

Governments all over the world have been fast to reply. Malaysia and Indonesia blocked access to Grok, whereas regulators within the U.Ok. and European Union opened investigations into potential violations of on-line security legal guidelines.

The U.Ok.’s on-line regulator, Ofcom, stated it might continue its investigation regardless of the newly introduced adjustments.

Within the U.S., California Lawyer Basic Rob Bonta introduced Wednesday that his workplace had launched its personal investigation into the issue.

In the meantime, as scrutiny of Grok has intensified, X quietly up to date its phrases of service to require that every one pending and future authorized circumstances involving the corporate be filed within the Fort Value division of the Northern District of Texas, the place one of many courtroom’s three judges is extensively seen as pleasant to the corporate.

Left-leaning watchdog Media Issues, a frequent critic of Musk’s X, stated it would leave the platform in response to the up to date phrases.