Grok deepfakes and SpaceX IPO: what French prosecutors are really telling the SEC

On May 7, 2026, the Paris prosecutor's office requested a formal investigation against X and Elon Musk. Nine counts, including complicity in the distribution of AI-generated child sexual abuse material via Grok and falsifying the operation of an automated system. The file is heavy.

The most unusual element, though, sits outside the French procedure. It was sent to Washington two months earlier, and it says something no regulator had ever stated publicly: an AI scandal can be a stock manipulation tool.

What Paris told the SEC

On March 17, 2026, the Paris prosecutor sent a formal note to two US agencies, the Department of Justice and the Securities and Exchange Commission. The wording, later quoted by Euronews and The Washington Post, deserves to be reproduced verbatim.

The controversy around Grok-generated sexually explicit deepfakes “may have been deliberately orchestrated to artificially boost the value of the companies X and xAI.” Deliberately orchestrated. To inflate the value of X and xAI.

Some precision is in order here. French prosecutors are alleging. The SEC and the DOJ are looking. Market manipulation has not been proven. What has been documented is the underlying material.

The documented facts

In late December 2025, xAI rolled out a Grok update that let users generate and edit images directly inside X. The Center for Countering Digital Hate started measuring.

Between December 29, 2025 and January 8, 2026, over eleven days, CCDH researchers estimated 3,002,712 sexualized images produced by Grok on the platform. Among them, 23,338 photorealistic images of sexualized children and 9,936 non-photorealistic equivalents, based on the NGO's statistical projections. 144 images were reported to the Internet Watch Foundation.

On February 3, 2026, the Paris prosecutor's office, backed by Europol, raided X's French offices. The same month, Elon Musk officially announced the merger between SpaceX and xAI at a combined valuation of $1.25 trillion.

On April 20, Musk and former X CEO Linda Yaccarino were summoned to a voluntary hearing. They did not show up. On May 7, the prosecutor escalated and requested a formal nine-count investigation.

Financial calendar in parallel. On May 20, 2026, SpaceX filed its preliminary S-1 with the SEC, with planned ticker SPCX on Nasdaq and Nasdaq Texas. Targeted valuation between $1.75 trillion and $2 trillion, with up to $75 billion to be raised. An all-time record. The document also details 2025 losses of $4.9 billion on $18 billion in revenue, and a project for orbital AI data centers carried by Starship and the merged xAI. Roadshow begins early June, pricing on the 11th, listing targeted for June 12, 2026.

Why Paris is talking about manipulation

The prosecutor's case does not rest on one isolated fact. It builds on a documented bundle of elements: a merger timing that follows the Grok scandal media peak by a few weeks, CCDH figures that look more like mass distribution than a one-off technical glitch, public promotion of the feature by Musk himself despite warnings, no increased moderation after the first alerts, and a refusal to respond to a judicial summons. None of this proves anything in isolation. Taken together, it was enough for a French prosecutor to cross the Atlantic with a note.

The angle is new because it steps out of the usual register. When people talk about AI scandals, they talk moderation, ethics, bias. When people talk market manipulation, they talk insider trading, false financial information, pump and dump. The Paris note bridges the two: it potentially recasts an AI content scandal as a financial communication instrument. That recasting is what is unprecedented.

What it says about AI regulation

On May 7, 2026, the same day French criminal escalation happened, the European Commission also confirmed that the full application of the AI Act would be pushed back to 2027–2028 via the Omnibus package. Dedicated AI regulation retreats at the exact moment the case that could have tested its utility moves into a French investigation. The coincidence is instructive.

Because the lever that bites here is not, in fact, AI-specific legislation. It is the classic toolkit of criminal and financial law: child protection, breach of the confidentiality of communications, falsification of automated systems and, on the US side, market manipulation and investor disclosure obligations. None of those texts was written for AI. They predate Grok. They apply to Grok.

It raises a real question, one we have the right to leave open: if ordinary law already bites, what exactly are we waiting for from the AI Act? Part of the answer lies in prevention rather than punishment. The other part is that a determined prosecutor in Paris has delivered in four months what years of European negotiation have not.

The real issue is speed

Pump and dump is hardly a new invention. Artificially inflating the value of an asset before selling it is one of the oldest playbooks in financial capitalism. What is new is the industrial speed at which the scandal itself, used as amplifier, can be manufactured.

Three million images in eleven days. No pre-AI tool made that volume possible. Without Grok, the same scenario would have required months and a visible, traceable budget. With Grok, it took a few seconds of compute per image, and it spread organically.

That is what the Paris prosecutor's note signals, between the lines, to the SEC. And US pressure just stepped up on another front: on May 19, 2026, the FTC began enforcing the Take It Down Act, which forces platforms to remove non-consensual intimate imagery, deepfakes included, within 48 hours of a report. More than $50,000 per violation. FTC chair Andrew Ferguson had sent prior formal warnings to Meta, Apple, Microsoft, TikTok, Reddit, Snapchat and X. X is explicitly named, and the French investigation keeps climbing in parallel. On June 12, if nothing changes, the largest IPO in history opens its first day of trading. US regulators will, or will not, read the Paris note.