More than half of all traffic flowing through Grok, Elon Musk’s flagship AI product, now comes from users requesting pornographic images, explicit videos, and erotic roleplay — and a new investigation reveals that xAI made this outcome a deliberate strategic choice, not an accidental failure it is scrambling to reverse. The company has simultaneously set aside $530 million to cover legal costs from that strategy while admitting internally that its engineers have found no reliable way to block child sexual abuse material generation without dismantling the explicit content business it is betting on.
The revelation, published June 25 by The Information and confirmed through secondary reporting by Engadget and multiple other outlets, reframes every legal proceeding, regulatory investigation, and government contract xAI currently holds: the company did not stumble into being the world’s largest AI adult content platform. It chose to be.
Grok Generated 10 Billion Images a Month While Rivals Walked Away
The scale of what xAI has built is significant. SpaceX IPO filings, cited in The Information’s reporting, show Grok produced 10 billion images and 2 billion videos per month during the first quarter of 2026. Two former xAI employees told The Information that the dominant use driving those numbers — accounting for well over half of all Grok traffic — is adult content: pornographic image generation, explicit video creation, adult roleplay chat, and large volumes of erotica requests.
Rivals chose differently. OpenAI, Anthropic, and Google have each prohibited explicit content generation across their consumer-facing products. That consensus created a gap, and xAI moved to fill it, investing engineering resources in expanding Grok’s image and video generation capabilities specifically in the directions its competitors declined to pursue.
The market position has produced measurable traffic — but not measurable competitive advantage. Grok’s overall web traffic fell 22% between January and May 2026, the steepest decline among any major AI platform tracked by Similarweb, even as Claude’s traffic grew 369% and Gemini’s grew 40% over the same period. Vital Knowledge analyst Adam Crisafulli called the strategy a move by a company that has “fallen further behind” its competitors, describing it as a drive to stay relevant as Grok trails on capability benchmarks.
The Coding Model Arbitrage: How Adult Content Spread Beyond Its Intended Channel
The adult content demand has spread beyond the product features xAI designed for it. Users discovered that routing explicit content generation requests through Grok’s coding-focused models costs less than using the primary consumer interface. An internal xAI analysis, described in The Information’s reporting, found that a significant share of requests to those coding models were for pornographic or nude images rather than anything related to software development.
This is not merely a curiosity. It reveals a structural feature of how content moderation is implemented inside the Grok product stack. Content restrictions at xAI are applied at the individual model level rather than at a unified inference gateway, meaning that a user who routes a request to a lower-priced model endpoint can, in practice, bypass restrictions designed for the primary consumer product. The pricing differential created a bypass — and users found it.
Why xAI Cannot Engineer Its Way Out of the CSAM Problem
The most consequential finding in The Information’s investigation is also the least visible from the outside: xAI engineers working on Grok have found no reliable technical fix that permits explicit adult content generation while blocking prompts designed to produce child sexual abuse material (CSAM).
This reflects a fundamental characteristic of how generative image models work. A model capable of producing explicit adult images has been trained on representations of human bodies in sexual contexts. The same underlying capability that produces compliant adult content can produce CSAM with a small shift in prompt context — a younger-described subject, a reference to school or family, a stated age. Filtering systems applied after generation can catch known patterns, but they cannot anticipate every variation, and users who understand the system will probe its edges. xAI’s engineers acknowledged internally that there was no easy fix for this dilemma, according to The Information.
Canada’s Office of the Privacy Commissioner reached a parallel conclusion in June 2026, finding that xAI’s remediation steps — which reportedly reduced unwanted sexual content violations by roughly half — still left the platform capable of generating and distributing non-consensual sexualized deepfakes. According to the OPC’s PIPEDA findings, halving the violation rate when the baseline was millions of images per month does not constitute an adequate safeguard.
The Center for Countering Digital Hate documented more than 23,000 sexualized images of children generated by Grok during an 11-day window spanning late December 2025 and early January 2026 — a period during which Grok was generating sexualized deepfakes at an estimated rate of more than 6,000 per hour.
Legal Proceedings Span Six Jurisdictions — and Keep Growing
xAI is currently under legal scrutiny in the United States, the European Union, the United Kingdom, France, India, Brazil, and Australia, with active proceedings in most of those jurisdictions.
In the U.S., a California class-action filed in March 2026 on behalf of three Tennessee teenagers alleges that xAI knowingly designed, marketed, and profited from Grok’s image generation capabilities while refusing to implement standard safety measures. The suit is one of the first to seek to hold an AI company directly liable for the production and distribution of AI-generated CSAM depicting identifiable minors. The city of Baltimore filed a separate suit in March 2026 alleging violations of consumer protection law. UK Labour Member of Parliament Jess Asato filed a High Court claim in London in June 2026, alleging xAI violated laws on misuse of private information and data protection after Grok was used to produce sexually explicit images and video depicting her without consent, according to Engadget.
In France, prosecutors launched a criminal investigation following raids on X’s Paris offices in February 2026. Investigators are examining whether viral deepfake controversies were deliberately amplified to boost xAI’s valuation ahead of the SpaceX merger and IPO, according to TechTimes reporting from April 2026. Elon Musk was summoned for a voluntary interview in Paris in April 2026.
SpaceX’s IPO prospectus, filed in June 2026, set aside $530 million to cover potential litigation losses tied to Grok’s image generation capabilities — the clearest signal yet that the legal exposure is material enough to require a formal financial provision in public company disclosures, as TechTimes reported.
Grok Also Serves the Pentagon: A Contradiction the Government Has Not Addressed
In December 2025, the U.S. Department of War announced a partnership with xAI to embed Grok into GenAI.mil, the department’s internal AI platform, at Impact Level 5 security clearance — the tier that permits handling of Controlled Unclassified Information. According to xAI’s government announcement, the agreement covers approximately 3 million military and civilian personnel. xAI also holds a $200 million ceiling contract with the Department of Defense and has made its models available to all federal agencies through the General Services Administration schedule, according to an ExecutiveGov report on the partnership.
Neither the Department of War nor xAI has publicly addressed whether Grok’s adult content strategy or the active CSAM-related legal proceedings create any conflict with those government partnerships. The gap between the platform’s dominant commercial use and its government deployment has grown more visible as the scale of the adult content business has become public.
What the Strategy Signals — and What It May Cost
A product whose dominant use case is explicit content generation faces compounding risks that a productivity or coding product does not. Brand association is the first: enterprise clients, corporate procurement teams, and the government agencies that are current xAI customers operate in professional environments where association with the world’s largest AI adult content platform is, at minimum, a reputational liability.
Recruitment is the second: not every engineer who joined xAI to work on frontier AI research is comfortable being assigned to build an anime-inspired adult roleplay avatar, as reports of internal friction confirm, according to Engadget’s coverage of The Information’s investigation.
The third and most durable risk is legal. The Take It Down Act, which took effect May 19, 2026, gives the Federal Trade Commission authority to fine covered platforms $53,088 per image not removed within 48 hours of a valid takedown request, as TechTimes reported when enforcement began. At the volume xAI is operating — billions of images per month — even a small noncompliance rate produces exposure that cannot be absorbed by a $530 million reserve.
The traffic data tells a story of a company that found a large, underserved demand and moved to capture it. Whether that demand can sustain a frontier AI company through the litigation, regulatory, enterprise, and talent consequences that come with it is a question the $230 billion valuation assigned to SpaceX at IPO has not yet been required to answer.
Frequently Asked Questions
What percentage of Grok’s traffic comes from adult content?
Two former xAI employees told The Information that well over half of all Grok traffic is driven by adult content, including pornographic image generation, explicit video creation, adult roleplay chat, and large-volume erotica requests. SpaceX’s IPO filings indicate Grok generated 10 billion images and 2 billion videos per month in the first quarter of 2026.
Does Grok have active government contracts, and does xAI’s adult content strategy affect them?
Yes. xAI holds a $200 million contract with the U.S. Department of Defense and is deployed on the Pentagon’s GenAI.mil platform for approximately 3 million military and civilian personnel at Impact Level 5 security clearance. Neither xAI nor the Department of War has publicly addressed whether the adult content strategy or the active CSAM-related lawsuits create any conflict with those government partnerships.
Why can’t xAI simply block child sexual abuse material while allowing adult content?
According to The Information’s reporting, xAI engineers found no reliable technical fix. Generative image models that can produce explicit adult content have been trained on sexual representations of human anatomy; the same underlying capability can produce child sexual abuse material when a prompt shifts the described subject to a minor. Filtering systems can reduce the volume but cannot eliminate it, and users who understand the system probe its edges. This is the architectural reason why content moderation in permissive generative AI systems cannot be treated as a solved engineering problem.
Is xAI facing criminal charges?
No criminal charges have been filed against xAI or Elon Musk as of June 26, 2026. However, French prosecutors opened a criminal investigation in February 2026 following raids on X’s Paris offices, examining whether deepfake-related controversies were deliberately amplified to inflate xAI’s valuation. Multiple civil lawsuits are active in California, Maryland, and the United Kingdom. The Take It Down Act, in effect since May 2026, carries criminal penalties of up to three years for violations involving minors.