Skip to main content
BrowseReadFor organisations
Legal · Public disclosure

AI and Algorithmic TransparencyDraft skeleton — CTO factual input required

Last updated: 2026-05-19

Contents (8 sections)
  1. Why this statement exists
  2. Where AI is used
  3. What it doesn't do
  4. Search and ranking
  5. Human oversight
  6. What we don't use AI for
  7. Provider transparency
  8. Reporting concerns

Why this statement exists

We use AI tools to help us run the Nexus directory. We think you should know what those tools do, what they don't do, and where a person — not a machine — makes the decisions that affect what you see.

This statement is voluntary today; the regulatory landscape will likely require something like it within the next 12-24 months. We are publishing it now because it is the right way to operate.

Where AI is used at Mens Network

Sherlock — Listing classification

We run an internal system called Sherlock. Its job is to read information submitted by organisations applying to list on Nexus, and to suggest a category for the listing (for example: SUPPORT, RECOVERY, COMMUNITY, ACTIVITY).

  • What it does: reads listing text submitted by the organisation, compares it against our category criteria, and proposes a category and confidence score.
  • What it does not do: decide whether a listing goes live. A human reviews every classification before it is published, particularly for listings entering SUPPORT or RECOVERY categories.
  • Models used: [CTO TO CONFIRM — Claude (Anthropic), GPT-4-class (OpenAI), routed via OpenRouter]. The choice of model for a given task is determined by our LLM Data Tiering Policy.
  • Data sent to providers: Organisation listing text as submitted to us. We do not send personal data of end-users (men looking for listings) to AI providers.
  • Provider locations: [CTO TO CONFIRM — typically US-based; transfers governed by SCCs and the LLM Data Tiering Policy].
  • Retention by providers: [CTO TO CONFIRM — typically 30 days for the API tier we use, per the relevant DPA].

What Sherlock does NOT do

  • Sherlock does not make decisions about which men see which listings.
  • Sherlock does not personalise the directory based on your behaviour.
  • Sherlock does not profile users or visitors.
  • Sherlock does not make legal, medical, financial, or safeguarding decisions about any person.

The Nexus search and ranking is deterministic — based on category, location, and listing attributes — not personalised by AI. Two people searching for the same thing in the same location see the same results.

Human oversight

Every classification decision affecting whether a listing is published, and every safeguarding-relevant decision (suspending an account, removing a listing for a reported safeguarding concern), is reviewed by a person.

If you believe an AI-supported decision has affected you and you want a human to review it, email privacy@mensnetwork.global with "AI review request" in the subject line. You have this right under Article 22 of the UK GDPR even though we do not believe our AI use produces significant decisions about you.

What we do not use AI for

  • We do not use AI to write replies to crisis-related messages.
  • We do not use AI to moderate SUPPORT or RECOVERY content automatically — a person reviews these.
  • We do not sell, share, or otherwise make available to AI providers any personal data of end-users without lawful basis and disclosure in our Privacy Notice.

Provider transparency reports

Where the AI providers we use publish transparency reports, we read them and update this statement when their position materially changes. We will update this statement when we change providers or when our use of AI materially expands.

Current providers' transparency frameworks (last reviewed 2026-05-19):

  • Anthropic — Acceptable Use Policy; model card / system card for each Claude model release.
  • OpenAI — Usage Policies; model cards.
  • OpenRouter — routes our requests to underlying providers; their policies apply to the routing layer.

Reporting concerns

If you have a concern about how we are using AI, or about a specific automated decision:

  • Email: privacy@mensnetwork.global
  • Right to human review: see "Human oversight" above
  • Complaints to the regulator: ICO at ico.org.uk — though as of 2026, no UK regulator specifically governs AI use in our category beyond the existing UK GDPR / Equality Act / OSA framework. We will update this if that changes.