1 day ago
5
The UK's accusation watchdog volition analyse reports that Elon Musk's AI chatbot, Grok, has been utilized to make intersexual imagery of children.
Grok was developed by Musk's xAI successful 2023, designed to beryllium a "truth-seeking" adjunct with a witty, rebellious personality.
Integrated into X, formerly Twitter, it uses real-time information from the level to make text, images and code.
But complaints person mounted that Grok was being utilized to make intersexual photos of existent women and children, and present the Information Commissioner's Office (ICO) is investigating.
Grok: 'The creators don't cognize however it works - oregon however to support it nether control'
The announcement comes connected the aforesaid time the X offices successful Paris were raided by French prosecutors examining akin allegations.
In a connection connected its website, the ICO confirmed it had opened a ceremonial probe into 2 X companies concerning their processing of idiosyncratic information successful narration to Grok, and the AI's imaginable to nutrient harmful sexualised representation and video content.
"We person taken this measurement pursuing reports that Grok has been utilized to make non‑consensual intersexual imagery of individuals, including children," the connection said.
"The reported instauration and circulation of specified contented raises superior concerns nether UK information extortion instrumentality and presents a hazard of important imaginable harm to the public."
William Malcolm of the ICO said the probe would probe whether X Internet Unlimited Company and xAI had complied with information extortion laws, and provided capable safeguards.
He said: "The reports astir Grok rise profoundly troubling questions astir however people's idiosyncratic information has been utilized to make intimate oregon sexualised images without their cognition oregon consent, and whether the indispensable safeguards were enactment successful spot to forestall this.
"Losing power of idiosyncratic information successful this mode tin origin contiguous and important harm. This is peculiarly the lawsuit wherever children are involved.
"Our relation is to code the information extortion concerns astatine the centre of this, portion recognising that different organisations besides person important responsibilities."
Ofcom besides investigating
Another regulatory watchdog, Ofcom, besides opened a ceremonial probe into X past period nether the UK's Online Safety Act, to find whether the steadfast was complying with its duties to support radical from amerciable content.
The European Commission launched a probe into Grok past period too, looking astatine whether it disseminates amerciable content, specified arsenic manipulated sexualised images, successful the EU.
The EU is 1 of a fig of authorities astir the satellite to person raised concerns astir Grok, with authorities officials successful Germany, Sweden, India, Japan, Malaysia, California, Indonesia and Philippines among those that person spoken out.
Mr Malcolm said the ICO was moving intimately with Ofcom and "international regulators".
Grok has restricted representation editing but yet to accidental where
In a abstracted statement, regulator Ofcom said - portion it was probing X - it was not investigating xAI, which provides the standalone Grok chatbot app.
Ofcom besides said its probe into X was inactive gathering grounds and warned the probe could instrumentality months.
On wherefore it was not investigating xAI, the connection said: "When we opened our probe into X, we said we were assessing whether we should besides analyse xAI, arsenic the supplier of the standalone Grok service.
"We proceed to request answers from xAI astir the risks it poses. We are examining whether to motorboat an probe into its compliance with the rules."
Ofcom said that due to the fact that of the mode the Online Safety Act relates to chatbots, it was presently incapable to analyse the instauration of amerciable images by the standalone Grok itself.
xAI said connected 14 January it had restricted representation editing for Grok AI users and blocked users, based connected their location, from generating images of radical successful revealing covering successful "jurisdictions wherever it's illegal". It is yet to place the countries wherever those restrictions apply.
xAI earlier said it had constricted the usage of Grok's representation procreation and editing features to paying subscribers only.
English (US) ·