An LLM-based assistant to query curated georeferenced datasets on femicides and LGBTIQ+ hate crimes. Uses generative AI with retrieval from trusted sources, ensuring traceability and contextualized answers.
Safety criteria— by category, in rubric order
Privacy & data · 20% of score
60% · 5/8 rated
✓Core content accessible without creating an account— Yes
✓No PII collected by default (name, email, IP, location)— Yes
~Privacy policy in plain language with a short summary— Partially
—User data not sold or shared with third parties— N/A
~Option to use pseudonym or alternative name— Partially
✕No auto-save of form content (shared device risk)⚠ CRITICAL— No
—Opt-in (not opt-out) consent — granular and revocable— N/A
—End-to-end encryption used where applicable— N/A
Safety by design · 20% of score
17% · 3/6 rated
✕Quick-exit button present (redirects discreetly to neutral page)⚠ CRITICAL— No
—Content warnings before sensitive or triggering content— N/A
—Disguised notifications or discreet app icon option— N/A
✕No "people you may know" feature that could expose users to abusers— No
~Two-factor authentication available— Partially
—Safe alternative contact options supported (distinct from personal accounts)Not assessed— Not assessed
Agency, consent & participatory design · 15% of score
25% · 4/7 rated
—User controls what information is public vs. private— N/A
✕Multiple content formats available (video, text, audio)— No
✓Self-paced use possible — no forced sequential flows— Yes
—User can opt out of tracking and cookies— N/A
✕Survivors co-designed AND co-iterated — not consulted once only— No
✕Tool does not force a single pathway (e.g. 'call police' only)— No
—Feedback mechanism exists for users to flag issues or suggest content— N/A
Accessibility, equity & inclusion · 15% of score
100% · 6/9 rated
✓Available in more than one language— Yes
✓Free to use — no paywall for core content— Yes
✓Mobile-friendly / responsive design— Yes
—WCAG 2.1 AA accessibility compliance documentedNot assessed— Not assessed
—Gender-neutral, non-stigmatising language throughout— N/A
✓Covers intersectional forms of GBV — not single-issue only— Yes
✓Works on low bandwidth / poor connectivity— Yes
✓Not geographically restricted in access— Yes
—Mitigates risks for users on shared or monitored devices— N/A
Transparency & accountability · 10% of score
40% · 5/6 rated
✓Creator / organisation clearly identified on the tool— Yes
~Last update date visible (signals active maintenance)— Partially
✕Scope and limitations clearly stated — including what the tool cannot do— No
✕Funding sources disclosed— No
~Methodology publicly documented— Partially
—Content aligned with global GBV/HP standards and reviewedNot assessed— Not assessed
Tone & trauma-informed UX · 10% of score
100% · 3/5 rated
✓No graphic or violent imagery — no sensationalised depictions of violence— Yes
—Uses "survivor" rather than "victim" language throughout— N/A
✓Warm, non-judgmental, validating tone throughout the interface— Yes
✓Visuals are empowering and gender-diverse — not stereotyping— Yes
—Does not minimise or hierarchise different forms of abuse— N/A
AI & algorithmic ethics · 5% of score
67% · 3/5 rated
✓Clearly states if and how AI is used in the tool— Yes
—Clearly states what AI is usedNot assessed— Not assessed
—Provides details of how user's inputs are processed/ saved/ shared with third parties.Not assessed— Not assessed
✓Chatbots / automated tools clearly identify themselves as non-human— Yes
✕Human escalation pathway available when AI cannot safely respond— No
Sustainability & feminist tech ethos · 5% of score
100% · 1/3 rated
✓Led or co-led by women, gender-diverse people, or Global Majority organisations— Yes
—Openly licensed — can be used, remixed or adapted by others— N/A
—Clearly states compliance with feminist values or trauma-informed design or survivor-centred approachNot assessed— Not assessed
See the Transparency page for how the weighted score and bands are calculated.