Astax Labs · AI Compliance
In force since 2 August 2026

Does your AI assistant tell people it’s AI?

Article 50 of the EU AI Act requires AI systems that interact with people to make clear they are AI. Most chat assistants still open with “Hi, how can we help?” and say nothing at all.

We assess your user-facing AI systems, tell you exactly what needs to change, verify the change once it’s live, and give you a dated evidence pack you can hand to procurement, an insurer, or a customer’s security review.

The fix itself is usually a settings change and takes about twenty minutes. The documentation is the part that takes time, and the part people actually ask you for.

The deadlines

Two dates that matter

DateObligation
2 Aug 2026
in force now
AI systems interacting with people must disclose that they are AI. Deepfake content must be disclosed.
2 Dec 2026
transitional deadline
Synthetic audio, image, video and text must be marked in a machine-readable format and detectable as AI-generated.

Engagements

What it costs

Disclosure assessment & fix

£299 · 48 hours
  • Assessment of your customer-facing chat assistant
  • Exact remediation steps for your platform
  • Verification once the change is live
  • Dated evidence pack, findings, before/after screenshots, scope reasoning

Full Article 50 readiness audit

from £1,200 · 5 days
  • Every user-facing AI system, not just the chat widget
  • Obligation-by-obligation position with scope reasoning
  • Prioritised gap list and remediation plan
  • Evidence pack, plus re-assessment before the December deadline

No charge

Check your site now

Put your address in below and we’ll tell you, right now, whether we can see a chat or AI assistant on your site and what it means for Article 50. No charge, no email required, no waiting.

We fetch your homepage once and read the page source. We do not sign in, submit anything, or store your page.

The practice

Inspect the assessment tooling

Astax Labs builds software and security tools. We publish the tooling we assess with: our open-source red-team scanner for large language models runs 29 probes across six OWASP LLM Top 10 categories and grades findings A to F. It is published under an MIT licence at github.com/Astaxxxx/llmscan, where you can read every probe before you engage us.

Our Article 50 methodology comes from primary research rather than vendor documentation. The public scanner repository includes the probe definitions and a sample report, so you can inspect the tooling directly.

Every engagement is delivered with a dated evidence pack: findings, scope reasoning, and before-and-after verification, written to be handed to procurement, an insurer or a customer's security review without further explanation.

Not legal advice. Astax Labs is not a law firm. Whether the Regulation applies to your business is a legal question, and where scope is uncertain our reports say so explicitly. What we provide is a technical assessment of what your systems actually disclose to users, and a documented record of any changes made.