Research · Sample frame live

Article 50 SaaS chatbot disclosure benchmark

Quick answer

Original dataset: a conservative review of public AI SaaS chatbot and agent surfaces after 2 August 2026. Version 1 contains 20 coded observations from a planned 50-product frame; “Unclear” means the public evidence was insufficient, not that disclosure is absent.

Reviewed 27 August 2026 · Preliminary technical reference, not legal advice.

Primary sources: EU AI Act (Regulation (EU) 2024/1689) — EUR-Lex · European Commission — AI Act overview · AI Act Service Desk — implementation timeline

Not legal advice. ActBrief prepares a preliminary technical readiness brief for legal review and enterprise diligence — not a conformity assessment, certification, or guarantee of compliance.

Status

Planned frame: 50 products · Coded v1: 20 · Pending expansion: 30. Methodology published 16 Aug 2026. Observational UX research only — not legal advice, not a compliance score for any vendor.

Version 1 result
0 explicit disclosures among 0 observations with a determinate result; 20 remain Unclear. Percentages exclude ActBrief (self) and exclude Unclear observations from the rate.

Why this research

Generic “what is Article 50?” posts are everywhere. A coded sample of live SaaS UIs after the application date is rare — and citeable. We publish the frame and rubric first so methodology is auditable before headline %.

Inclusion rules

Coding rubric

disclosure_present — Y / N / Unclear
Y only if a natural person can see explicit AI / assistant / bot interaction notice on the observed surface before or during interaction.
placement — pre_chat | in_chrome | first_message | footer_only | privacy_only | outbound_signature | none | other
Code the primary notice location. footer_only / privacy_only alone → still disclosure_present may be Y for that surface, but flagged as weak placement for analysis.
wording_mentions_ai — Y / N / Unclear
Y if text mentions AI, artificial intelligence, assistant, bot, or equivalent clear wording.
persistent — Y / N / Unclear / NA
Y if notice remains visible for the session; N if dismissible toast only; NA if no disclosure.
screenshot_date — ISO date
Date the public surface was observed. Store screenshots offline; do not publish private account data.

Coded v1 sample (20 of 50 planned products)

Full sheet: download CSV · download JSON.

#ProductVerticalStatus
01Intercom FinsupportCoded
02Zendesk AIsupportCoded
03Freshdesk FreddysupportCoded
04HubSpot BreezesalesCoded
05Salesforce AgentforcesalesCoded
06Drift / SalesloftsalesCoded
07AdasupportCoded
08ForethoughtsupportCoded
09CapacitysupportCoded
10SierrasupportCoded
11DecagonsupportCoded
12GorgiassupportCoded
13CrispsupportCoded
14TidiosupportCoded
15LiveChat / TextsupportCoded
16AshbyhrCoded
17GreenhousehrCoded
18LeverhrCoded
19HireVuehrCoded
20Paradox OliviahrCoded

How this dataset should be cited

“ActBrief Article 50 Chatbot Disclosure Benchmark, version 1.0, reviewed 27 August 2026.” The dataset is available under CC BY 4.0; cite the observation URL and date for any individual row.

Expansion plan

  1. Expand the coded sample from 20 to the full 50-product frame.
  2. Report explicit disclosure, placement, persistence, and Unclear counts separately.
  3. Keep screenshots offline; publish only public URLs, dates, coding fields, and notes.
  4. Retain sampling-bias and non-legal limitations on every version.

Related implementation

Chatbot disclosure · Placement · AI agent disclosure · Evidence log · ActBrief methodology

Official sources

Continue in this cluster

Scan your own chatbot while we code the sample

Free readiness brief with Art. 50 signals and evidence gaps. Email optional.

Get a free readiness brief for your product