The EU AI Act's transparency rules have applied since 2 August 2026. A one-off observation cannot show whether selected deployment conditions later changed.
Regula Article 50 Monitor is a planned independent service that repeatedly observes selected chatbot disclosure and AI content label surfaces, records what was observed, by which method and source version, and routes uncertain or contextual findings to named human review. It does not determine compliance, truth, certification or regulator acceptance.
Why this is being tested
The problem with a point-in-time check
One-off Article 50 checkers exist, and some are free. They provide a point-in-time observation. The question is whether selected disclosure surfaces or source versions changed after review. Regula would record selected observations over time, identify changes, and route contextual or uncertain findings to a named reviewer. Official guidance requires contextual assessment of whether information is clear, distinguishable, accessible and timely. Some requirements cannot be established by a technical surface check alone. Regula does not determine compliance, truth, certification or regulator acceptance.
Proposed workflow
From observation to review
Next step
Register interest
Founding places
Founding access is limited to 20 places. A founding place costs a GBP 29 deposit, credited as your first month at launch. The planned price is GBP 29 per month. Your deposit is refundable in full at any time before launch, and refunded automatically if we have not launched by 30 September 2026.
Founding checkout is temporarily unavailable.
Show interest
ReferenceRead the regulatory background
What Article 50 requires, in broad terms
- Systems that interact directly with people must make clear that the person is dealing with AI, unless that is already obvious (Article 50(1)).
- Providers of generative AI must ensure synthetic audio, image, video and text output is marked as artificially generated in a machine readable way (Article 50(2)).
- Deployers must visibly label deep fakes, and disclose AI generated text published to inform the public on matters of public interest (Article 50(4)).
- People exposed to emotion recognition or biometric categorisation systems must be informed (Article 50(3)).
- Article 99 sets a general maximum for Article 50 infringements of EUR 15 million or 3 percent of worldwide annual turnover, whichever is higher. For SMEs, including start-ups, the applicable maximum is the lower of the amount or percentage.
- These duties have applied since 2 August 2026. They are not limited to high-risk systems. SME status alone does not exempt an otherwise in-scope actor; role and use still require assessment. Open-source licensing does not exempt an otherwise in-scope system. A non-EU organisation can be in scope where Article 2's role and EU nexus tests are met.
- For providers of qualifying generative AI systems placed on the market before 2 August 2026, the Article 50(2) marking and detection obligations apply from 2 December 2026. This limited transition does not create a general grace period.
- Outputs generated before 2 August 2026 generally do not require retroactive marking under Article 50(2). Public-interest text generated before that date but published on or after it must still carry the required Article 50(4) disclosure.
- The Commission published Article 50 guidance materials on 20 July 2026. Official sources currently conflict on formal adoption and applicability; obtain current qualified advice on their status and use. The regulation text governs.