Grok (xAI) and GDPR / AI Act: a tool to handle with caution

Grok (xAI) and GDPR / AI Act: a tool to handle with caution

In brief

  • Grok, the AI from xAI (Elon Musk) integrated into X, has a high-risk profile. The Irish DPC opened an investigation in April 2025 on training from public posts from Europeans, and a second investigation in early 2026 on image generation. Data sharing for training is enabled by default on X, with an opt-out that is not easily visible. For professional use with personal data, the utmost caution is required.

Overview and data processed

Grok is a family of generative AI models developed by xAI, Elon Musk's company, integrated into the X platform (formerly Twitter). Its claimed advantage—access to billions of X posts—is precisely what crystallises regulatory concerns.

Grok processes prompts and responses, but also, as a model, trains on public posts from X users. On X, permission to use data to train Grok is enabled by default and opt-out is not easily accessible. Since late 2024, Grok also automatically generates information about users. US publisher (CLOUD Act).

Risks for your business

Beyond the usual risks, Grok combines reputational and regulatory risk: ongoing investigations, controversies over image generation (including a DPC investigation in early 2026 on images of a sexual nature), and contested legal basis for training. Associating your brand with unregulated professional use of Grok is risky.

Main GDPR risks:

  • Contested legal basis for training (legitimate interest vs consent), DPC investigation ongoing.
  • Insufficient transparency and right to object (opt-out option not easily visible).
  • Transfer outside the EU (CLOUD Act).
  • Processing of third-party data from public posts.

GDPR obligations

  1. If use is considered: favour an API/Enterprise offering governed by a DPA, not the public offering on X.
  2. Disable data sharing for training on the accounts concerned.
  3. Record in the record of processing activities; DPIA almost always required given the risk.
  4. Govern transfers (SCC); inform data subjects.
  5. Involve the DPO and set out a clear position in the AI charter.

AI Act obligations

Grok is a GPAI model (provider obligations since 2 August 2025). For deployers: transparency (Article 50), human oversight, AI literacy. Image/video generation also raises obligations for labelling generated or manipulated content. See the AI Act guide.

Cybersecurity and best practices

Cyber risks: leakage via prompts and via integration into X, generation of problematic content that can be exploited against the brand, shadow AI. To be factored in for NIS 2 entities.

Best practices: by default, advise against Grok for any processing of personal data in the enterprise; if used, governed offering + DPA, training opt-out, DPIA; monitor the outcome of DPC investigations before any deployment; explicit AI charter, training, AI lead, record.

Compliance checklist:

  • Documented decision on use/non-use of Grok
  • If used: governed offering + DPA, training sharing disabled
  • DPIA carried out (high risk), transfers outside the EU governed (SCC)
  • Position set out in the AI charter
  • Article 50 transparency and content labelling, teams trained, DPO involved

Sanctions, conclusion and recommendations

Sanctions (two cumulative regimes): GDPR up to €20m or 4% of total worldwide annual turnover; AI Act up to €35m or 7% for prohibited practices, and €15m or 3% for most other infringements. In the event of a breach, X's European entity faces a fine of up to 4% of its worldwide turnover; the user company remains fully liable under the GDPR.

Recommendation: Grok is, to date, one of the most exposed assistants from a regulatory standpoint. Our recommendation: do not use it to process personal data in a professional context until the DPC investigations are clarified, and set out an explicit position in the AI charter. If use is essential, it must be strictly governed (professional offering, DPA, DPIA, training opt-out). Compare with other assistants: ChatGPT, Meta AI, DeepSeek.

Take action

Assess the GDPR and AI Act risk level of your use of Grok with the Data Comply One self-assessment.

Discover the AI Act software →Test my compliance for free
Compare this tool with the 70 AIs analysed — hosting, training, DPA and verdict.See the full comparison →

Frequently asked questions

Is Grok GDPR-compliant?

Its risk profile is high: training on public posts from Europeans is the subject of an investigation by the Irish authority (DPC) since April 2025, and the legal basis is contested. Caution is required for any professional use with personal data.

Does Grok use my data for training?

On X, data sharing to train Grok is enabled by default, with an opt-out option that is not easily visible. Check and disable this setting.

Can I use Grok in the enterprise?

Not recommended by default for personal data until investigations are clarified. If necessary, only via a governed offering with DPA and DPIA.

Is Grok covered by the AI Act?

Yes, as a GPAI; image/video generation adds obligations for labelling generated or manipulated content.