An AI feature became a legal crisis. Guardrails are not optional.
The UK regulator's probe into X shows what happens when generative AI ships without limits. Every business adding AI should take note.
By the Xuru Technologies team · 13 January 2026
Key takeaways
- Ofcom opened a formal investigation into X under the Online Safety Act on 12 January 2026.
- It concerns sexualised images, including of real people, created with the Grok AI tool.
- Businesses using AI must limit what it can generate, monitor outputs and respond fast to misuse.
Generative AI is powerful and easy to misuse. When safeguards are weak, the damage lands on real people and, eventually, on the company that deployed it.
What happened
Ofcom announced a formal investigation into whether X met its duties to protect UK users from illegal content linked to Grok-generated images. The Register and the Society for Computers and Law covered the legal context. Under the Act, Ofcom can impose significant fines.
What it means for businesses
- You own your AI's output. Regulators look at the company that offers the feature.
- Reputation moves faster than law. One viral misuse can define a brand.
- Safety by design wins. Narrow tools with clear limits are safer than open-ended ones.
Safe AI checklist
| Control | Purpose |
|---|---|
| Limit the AI to your use case | A booking assistant should only discuss bookings. |
| Block image generation of real people | Prevents deepfakes and abuse. |
| Log and review conversations | Spot misuse early. |
| Easy reporting and human handoff | Users can flag problems and reach a person. |
Frequently asked questions
Why is Ofcom investigating X?To assess whether X complied with UK Online Safety Act duties after Grok was used to create sexualised images, including of real people.
To assess whether X complied with UK Online Safety Act duties after Grok was used to create sexualised images, including of real people.
Does the Online Safety Act apply to my business?It applies mainly to services that host user content or search. If your site lets users post or share, take advice on your duties.
It applies mainly to services that host user content or search. If your site lets users post or share, take advice on your duties.
Is a website AI assistant risky?Low risk when it is limited to your own content, cannot generate images and hands sensitive topics to staff.
Low risk when it is limited to your own content, cannot generate images and hands sensitive topics to staff.
AI assistants with guardrails built in.
Narrow, grounded assistants that stay on topic, log activity and hand off to people.
Websites, payments, digital marketing, AI automation, consultancy and practical technology training — for businesses and aspiring professionals worldwide.
Saturday – Sunday: closed