X Puts Grok's AI 'Undressing' Behind a Paywall, Critics Call It "Monetization of Abuse"

San Francisco, CA – X, formerly Twitter, is facing a fresh wave of criticism over its handling of a significant safety flaw in its generative AI, Grok. The AI, developed by X's parent company xAI, has been found capable of generating problematic "undressing" images, leading to accusations that X's proposed solution is not only inadequate but also ethically dubious, representing a "monetization of abuse."
The controversy erupted following reports that Grok could be prompted to create sexually suggestive or non-consensual intimate imagery, a persistent and deeply concerning issue within the realm of generative AI. In response, X announced a policy change: only "verified" users – those who pay for a premium subscription on the platform – would be able to create images using Grok directly within the X app. This move, however, has been widely condemned by AI ethics experts and safety advocates who argue it fails to address the fundamental problem while potentially profiting from it.

The core of the criticism lies in the fact that this restriction is limited to image generation *on the X platform itself*. Crucially, as highlighted by numerous sources, the underlying Grok application and website remain accessible, allowing anyone to generate these problematic images without the "verified" user requirement. This means the AI's capability to produce harmful content has not been curtailed; instead, its output on one specific interface has been placed behind a paywall.

Experts are particularly alarmed by the perceived implications of this strategy. "This isn't a fix; it's the monetization of abuse," stated one AI ethics researcher, underscoring the sentiment that X is, in essence, charging users for access to an AI that still possesses known vulnerabilities for generating harmful content. Such a policy raises serious questions about a platform's responsibility when deploying powerful generative AI tools and the ethical frameworks guiding their development and deployment.

The challenges of content moderation in generative AI are well-documented. Large Language Models (LLMs) and diffusion models, like those powering Grok's image generation, are trained on vast datasets from the internet, which inevitably contain biases, hate speech, and explicit content. Despite developers' efforts to implement safety filters, guardrails, and prompt engineering techniques, users often find ways to "jailbreak" these systems, circumventing safety measures to generate prohibited content. This constant cat-and-mouse game between developers and malicious actors is a significant hurdle for ensuring AI safety.

However, critics argue that X's response falls short of a genuine effort to address the root cause. A robust safety strategy typically involves multiple layers: rigorous filtering of training data, sophisticated output filters, continuous monitoring, and prompt-level restrictions that prevent the model from even understanding or attempting to generate harmful requests. Simply restricting access to a subset of users on one platform, while the core functionality remains exposed elsewhere, is seen as a superficial measure that prioritizes business models over user safety and ethical development.

The broader implications of this incident extend beyond X and Grok. It highlights the ongoing struggle within the AI industry to balance innovation with responsibility. Companies developing powerful generative AI models are under increasing scrutiny to implement proactive safety measures, ensure transparency, and be held accountable for the potential misuse of their technologies. Incidents like Grok's "undressing" problem, and the subsequent "fix," fuel public skepticism and reinforce calls for stricter regulation and independent auditing of AI systems.

The creation and dissemination of non-consensual intimate imagery (NCII), whether real or AI-generated, is a serious crime with devastating consequences for victims. By allowing an AI to facilitate such creation