Malaysia, Indonesia block Musk’s Grok as backlash grows over AI images

Malaysia and Indonesia have become the first countries to block access to Grok, the artificial intelligence chatbot developed by Elon Musk’s startup xAI, after authorities said the tool was being misused to generate sexually explicit and non-consensual images of women and children.

The unprecedented moves by the two Southeast Asian nations mark a sharp escalation in global scrutiny of generative AI tools and come amid mounting criticism of Grok across Europe, Britain and India. Regulators and campaigners say existing safeguards have failed to prevent the creation and spread of realistic sexualised “deepfake” images, raising concerns over privacy, safety and human rights.

Indonesia temporarily blocked access to Grok on Saturday, followed by Malaysia on Sunday, after both governments said the chatbot lacked effective controls to stop abuse.

“The government sees non-consensual sexual deepfakes as a serious violation of human rights, dignity and the safety of citizens in the digital space,” Indonesia’s Communication and Digital Affairs Minister Meutya Hafid said in a statement.

Initial findings showed Grok did not have adequate safeguards to prevent users from creating and distributing fake pornographic content using real photographs of Indonesian residents, Alexander Sabar, director general of digital space supervision, said. He warned that such manipulation risked violating privacy and image rights, causing psychological, social and reputational harm.

In Malaysia, the Communications and Multimedia Commission ordered a temporary restriction after what it described as “repeated misuse” of Grok to generate obscene and non-consensual manipulated images, including content involving minors.

“The restriction is imposed as a preventive and proportionate measure while legal and regulatory processes are ongoing,” the regulator said, adding that access would remain blocked until effective safeguards were put in place.

The backlash erupted after Grok rolled out an image editing feature that allowed users to alter existing photos using simple text prompts. Critics said the tool could be used to digitally undress women, sexualise images without consent, and even manipulate photos of children.

Users reported prompts such as “put her in a bikini” or “remove her clothes” producing sexualised images, often based on real photographs shared online. In several cases, images continued to circulate even after complaints were raised.

The controversy intensified following a fatal shooting in Minneapolis earlier this month, after manipulated images of the female victim — including sexualised versions — were generated using Grok and shared online, according to campaigners cited by AFP.

Amid the global backlash, Grok last week restricted its image generation and editing features to paying subscribers on Musk’s social media platform X, formerly Twitter. The company said users would need a premium subscription to access the tools.

However, European officials and safety advocates dismissed the move, saying it did not address the core problem.

“That simply turns an AI feature that allows the creation of unlawful images into a premium service,” a spokesperson for British Prime Minister Keir Starmer said, calling the change “insulting” to victims of misogyny and sexual violence.

EU digital affairs spokesman Thomas Regnier said restricting access did not resolve the issue. “Paid subscription or non-paid subscription — we don’t want to see such images. It’s as simple as that,” he told reporters.

Grok has acknowledged “lapses in safeguards” and said it was urgently fixing flaws, adding that child sexual abuse material is illegal and prohibited. In one instance, the chatbot issued an apology for generating a sexualised image of two underage girls based on a user prompt.

When contacted by AFP for comment on several occasions, xAI responded with an automated message: “Legacy Media Lies.”

Beyond Southeast Asia, regulators across Europe and Asia have stepped up pressure on Musk’s companies.

The European Commission, acting as the EU’s digital watchdog, has ordered X to retain all internal documents and data related to Grok until the end of 2026 as part of its inquiry. French prosecutors have expanded an investigation into X to include allegations that Grok was being used to generate and disseminate child pornography.

Britain’s media regulator Ofcom said it had made “urgent contact” with X and xAI and was conducting an expedited assessment to determine whether a formal investigation is warranted.

India has also directed X to remove sexualised content, clamp down on offending users and submit an “action taken report,” warning of legal consequences if it fails to comply.

Malaysia and Indonesia’s decision to block the tool entirely sets them apart from other jurisdictions, which have so far relied on investigations, takedown demands and regulatory warnings.

Launched in 2023, Grok is an AI chatbot developed by xAI and integrated into X. It allows users to ask questions, generate text responses and, more recently, create and edit images directly within posts and replies.

Last year, xAI added an image generation feature known as “Grok Imagine,” which included a so-called “spicy mode” capable of producing adult content. Critics say that feature, combined with weak safeguards, made the tool especially vulnerable to abuse.

While Musk has said anyone using Grok to create illegal content will face the same consequences as uploading such material directly, regulators argue that responsibility also lies with platform design and governance.

“Access restrictions alone aren’t a comprehensive safeguard,” said Cliff Steinhauer of the US-based National Cybersecurity Alliance. “Meaningful user protection needs to be grounded in how these tools are designed and governed.”

As governments grapple with the rapid spread of generative AI, Malaysia and Indonesia’s blocks signal a tougher approach — and raise the prospect that other countries may follow if safeguards fail to keep pace with misuse.

- with inputs from AFP, AP