ChatGPT against Grok: the assistant France uses most against the one Brussels put under investigation

ChatGPT has 662 million monthly active users worldwide, still captures 46% of the market for AI assistants, and has just filed to go public on a valuation of $852 billion. Grok is the third most used chatbot in France — just behind ChatGPT and Gemini — and has been tied since February 2026 to Elon Musk’s SpaceX empire, a combined entity valued at $1,250 billion. Three weeks before that merger, the European Commission opened a formal investigation into it over a wave of mass-generated sexual deepfakes, some of them depicting minors. Here is the match as of 7 July 2026 — figures, lawsuits and NGO reports included, without the marketing varnish.

---

The verdict, in short

Across the seven criteria in this comparison, ChatGPT wins 4-2, with one tie. It wins comfortably on real-world adoption, on price, on child safety and on the European regulatory file. But Grok is not beaten everywhere: on real-time access to the news and on its deliberately less filtered tone, it keeps a genuine lead.

---

Two philosophies, one race to a trillion

ChatGPT was born on 30 November 2022, the flagship product of OpenAI, founded by Sam Altman and a handful of researchers back in 2015. Its financial trajectory turned vertiginous in 2026: a $122 billion funding round closed on 31 March, led by Amazon ($50B), SoftBank ($30B) and Nvidia ($30B), taking the post-money valuation to $852 billion. On 8 June 2026, OpenAI filed to go public, targeting around $1,000 billion at listing. The company now claims 662 million monthly active users, up from 533 million last December, and still held 46% of the global AI assistant market in May 2026 — down, and below 50% for the first time, as Google Gemini (28%) and Claude (10%) break through.

Grok comes from somewhere else. Launched in late 2023 by xAI, the company founded by Elon Musk, it was built on the opposite positioning: a « less censored » AI, natively integrated into the X social network, with direct access to the real-time news feed. Its financial trajectory is just as spectacular: xAI’s valuation went from $24 billion in late 2024 to $230-250 billion in early 2026, before the company merged with SpaceX on 2 February 2026, creating a combined entity valued at $1,250 billion — an operation largely underwritten by projects not yet delivered. In France, Grok is the third most used AI app, behind ChatGPT and Gemini, according to Sensor Tower’s « State of AI » report published in late June 2026.

Two philosophies, then: one aiming at mass adoption and enterprise integration, the other at speed, breaking news, and an audience that values the absence of filters.

---

What each one actually does

ChatGPT long ago outgrew the simple chatbot. OpenAI has stacked on image and video generation (Sora), deep research, persistent memory across conversations, agents able to carry out tasks, and more recently openly commercial ambitions: built-in shopping, advertising, and an assistant known internally as « Aria » that is gradually turning ChatGPT into a commerce platform. The tool targets a general audience: consumers, professionals, developers (through Codex) and large enterprises.

Grok bets on different ground: freshness of information and native integration with X, which lets it analyse posts, images and trends in real time — a real advantage for following the news, where ChatGPT depends on conventional web search. xAI has also built Grok Imagine (image and video generation) and AI « companions » with personalities — including Ani, an animated companion with a « gothic » tone whose most advanced mode unlocks sexually explicit exchanges. One documented problem: the app was rated 12+ on the App Store despite that feature, which fed much of the concern detailed below.

---

The price, without the small print

At every comparable tier, ChatGPT is cheaper: $20 against $30 for individual entry, $200 against $300 at the top. Grok partly compensates through the X Premium+ bundle, which adds the platform’s benefits (less advertising, creator monetisation) — but that is not a pure AI subscription, it is a platform subscription that includes AI.

---

The angle most comparisons leave out: child safety

This is where this comparison parts company with the rest of the specialist press, which mostly compares answer quality and speed. On child safety and the production of non-consensual sexualised content, the file is heavy — and very unevenly distributed between the two tools.

The wave of sexualised deepfakes generated through Grok is extensively documented: a British MP, Jess Asato, is suing xAI after non-consensual sexualised images of her were generated; another complainant, Ashley St. Clair, says she was depicted as a minor in an image generated by Grok; the city of Baltimore has also begun proceedings. The Center for Countering Digital Hate put the number of sexualised images produced and published on X at around 3 million in just 11 days at the turn of 2025-2026. Faced with the outcry, xAI restricted the tool to paying users on 9 January 2026, then added further restrictions on 14 January — but a test run by Reuters on 3 February showed the tool still produced, in most cases, sexualised images of real people without their consent.

The file has grown heavier since. Common Sense Media, the US reference for digital safety assessment for young people, published a report on 27 January 2026 concluding that Grok does not correctly identify underage users, that its « Kids Mode » lets explicit content through, and that the tool handles signals of psychological distress dangerously. The head of the study puts it bluntly: Grok is « among the worst » chatbots the organisation has ever evaluated. The California Attorney General has opened an investigation into whether xAI broke US law. In France, a senator put a question to the government titled « Stop Grok », demanding a ban on « undressing » features; the government replied that it was « fully mobilised » on the problem. At European level, after a first fine of 120 million euros imposed on X on 5 December 2025 for breaching the Digital Services Act, the European Commission opened a new formal investigation on 26 January 2026 centred on Grok — with a potential fine of up to 6% of X’s worldwide revenue. Arcom, the French audiovisual and digital regulator, publicly welcomed the opening of that investigation.

---

OpenAI is no choirboy either

The picture would not be honest if it stopped there. OpenAI faces a serious file of its own: the family of Adam Raine, a 16-year-old American who died by suicide, is suing the company, saying ChatGPT gave him dangerous guidance — going as far as analysing a photograph of a noose and confirming it could « potentially suspend a human being ». OpenAI’s defence invokes « inappropriate » use of the tool.

The timing of OpenAI’s communications did not help its case: on 14 October 2025, barely two months after that lawsuit was filed, Sam Altman announced that ChatGPT would allow, from December, the generation of « erotica » for verified adults — in the name of « treating adults like adults ». The announcement drew a wave of criticism on a scale Altman himself admitted he had underestimated. Another telling fact: while investigating Grok, Le Recul tested AI image generation through ChatGPT and obtained a sexualised visual bordering on explicit, without using the image of a real person. The difference with the cases documented around Grok is real — no real person targeted, no public distribution — but it is a reminder that the risk is not confined to one vendor.

One substantial distinction separates the two files nonetheless: unlike Grok, ChatGPT has to date been the subject of no NGO report ranking it « among the worst » for child safety, and of no regulatory fine already handed down. The CNIL has opened a compliance procedure on ChatGPT after several GDPR complaints, but no European authority has yet issued a decision — a very different situation from a 120 million euro fine already imposed.

---

Regulatory compliance: two files, not the same weight

This is the most structural criterion in this comparison, and the one that explains most of the gap in the score.

Grok/X’s file is already settled on one point: 120 million euros in fines, imposed on 5 December 2025 for breaching the Digital Services Act. The new investigation opened on 26 January 2026 concerns specifically Grok’s image generation capabilities and X’s recommendation systems, suspected of amplifying illegal content — including content that may constitute child sexual abuse material, according to the Commission. If the breaches are confirmed, the potential fine climbs to 6% of X’s worldwide revenue, an amount on an entirely different scale from the penalty already imposed.

ChatGPT’s file, by contrast, remains open but unsettled: the CNIL has received several complaints and is conducting checks on OpenAI’s GDPR compliance (transparency about training data, the right to object), with no European authority having issued a decision either way to date. The European AI Act, a key milestone of which imposes reinforced safety obligations from 2 August 2026 on any AI interacting with children, will mechanically put both companies under the microscope — but Grok arrives there with a fine already paid and an investigation open, where ChatGPT arrives with a file still under examination.

---

The real limits, not hidden

ChatGPT is not free of structural criticism: its drift towards a commercial « super-app » (built-in shopping, advertising, transactional agents) raises the question of what becomes of an assistant originally designed to answer questions. Its decision to open up adult content, taken in a sensitive legal context, remains a questionable communications choice. And its market-share lead is already eroding against Gemini, whose progress is spectacular.

Grok, for its part, is not reducible to its safety scandal: its real-time access to X remains a genuine advantage for following breaking news, and its less filtered tone appeals to a specific audience that finds ChatGPT excessively cautious. But that strength and that flaw are two sides of the same coin: it is precisely the intent to impose as few guardrails as possible that allowed the wave of deepfakes documented above. A chatbot cannot claim the absence of filters as a selling point and then be surprised that the absence of filters does damage.

---

What to remember

• Overall winner in this comparison: ChatGPT (4 criteria out of 7, one tie). Real-world adoption, price, child safety and the European regulatory file tip the balance.
• Grok keeps two real strengths: instant access to the X feed, and a deliberately unfiltered tone — a genuine argument for part of its audience.
• Child safety is the clearest gap: Grok is called « among the worst » by Common Sense Media (27 January 2026); ChatGPT faces a serious lawsuit (the Raine case) but no comparable report.
• The regulatory file weighs heavily against Grok: a €120 million DSA fine already imposed, a new European investigation under way, a senator’s question in France. On the ChatGPT side, a CNIL procedure remains open, with no decision to date.
• Neither is an isolated case: Le Recul’s own test showed that ChatGPT too could generate a sexualised image bordering on explicit.
• The figure to remember: 6%. That is the share of X’s worldwide revenue the European Commission could take in fines if its investigation into Grok, opened on 26 January 2026, concludes that there were systemic breaches of the Digital Services Act.

For more of this kind of work, see our other AI comparisons and our tested promises.