Open-source intelligence
JunkipediaMedia CloudCase study
Mapping the Matryoshka network — from a single AI-generated video to the full campaign infrastructure.
Method · Two tracks, one cascade
Two tracks. One cascade.
AI Shield builds capacity on two levels at once. Track A ("Expert") develops deep analytical expertise in 12–15 analysts from AIISA and partner think tanks. Track B ("Multiplier") trains 30–36 staff from independent media and watchdog CSOs. The decisive innovation: Track A graduates become Track B trainers — they run the workshops themselves, without CSIN in the room. If AIISA's trainers can train the next cohort, the transfer worked.
Build the analytical core
- 12–15 analysts from AIISA and partner think tanks (target: APRI Armenia, Regional Studies Center, Regional Centre for Democracy and Security)
- 54 hours of training (30 in-person in Yerevan + 24 online)
- Every participant analyses a real disinformation campaign and writes a report
- Pre/post test: ≥ 30% competency gain
- Certification by 31 March 2027
Cascade through media and CSOs
- 30–36 media and CSO staff; target organisations include CivilNet, EVN Report, Hetq, Media Initiatives Center, Union of Informed Citizens / FIP.am, Public Journalism Club, Transparency International Armenia, FOICA, Helsinki Committee
- 3 workshop cycles × 10–12 participants
- Delivered by AIISA trainers (Track A alumni) — CSIN provides online mentoring only
- Measurement: ≥ 25% competency gain, workshop rating ≥ 4/5
If AIISA's trainers can train the next cohort without us in the room, the transfer worked.
Timeline · June 2026 — May 2027
Twelve months. Four activities. One trajectory.
Mapping the Armenian disinformation landscape
Three days of high-intensity diagnostics with 15 analysts from AIISA and invited think tanks. Poland's road from target to exporter of resilience (2014–2026). Narrative parallels between the Armenian 'Ukrainisation' frame and the 'NATO means war' frame deployed against Poland. Mapping the actors: Russia (Matryoshka, Storm-1516), Azerbaijan, Iran (IRGC). Survey of AI tools and curriculum design for Activity 2.
Training 12–15 analysts as trainers
In-person block in Yerevan, 5 days × 6 hours: AI-powered OSINT (Junkipedia, Media Cloud); narrative analysis and coordination detection on Storm-1516 (Botometer, Bot Sentinel); deepfake detection (InVID WeVerify, FotoForensics, Hive, TrueMedia.org); social network analysis (Gephi, NodeXL); practical integration and trainer preparation. Online block, 6 weeks × 4 hours: each participant picks a real campaign targeting Armenia, analyses it and writes a report, mentored by CSIN experts.
3 workshop cycles run autonomously by AIISA
Three 16-hour cycles (10–12 participants each) for staff from CSOs and independent media. AIISA trainers, certified in Activity 2, lead the workshops; CSIN only provides online mentoring. This is the empirical evidence of transfer — not a declaration, but a measurable demonstration that the partner can train the next cohort on its own.
Open products: Toolkit, Report, Panel
Open-source 'AI Shield' Toolkit — ~60 pages, three languages (PL/EN/AM), CC BY-NC 4.0. Five chapters: typologies of coordinated information operations, Poland's road from target to exporter, step-by-step tooling, Armenian case studies, action plan for newsrooms and CSOs. Analytical report distributed to Armenia's MFA, parliamentary committees, the EU Delegation. Expert panel in mid-May 2027 — hybrid format, minimum 50 attendees.
Toolkit · Four families, all free or freemium
Four tool families. All free or freemium.
No cost barrier to continued use after the project closes. The Toolkit documents each tool step by step, in three languages.
NLP & coordination detection
BotometerBot SentinelCase study
Analysing Storm-1516: account-synchronisation patterns, narratives, channels.
Deepfake detection
InVID WeVerifyFotoForensicsHive ModerationTrueMedia.orgCase study
The first audio deepfake of PM Pashinyan — the full verification trail.
Social network analysis
GephiNodeXLCase study
Visualising the coordinated-account network in the 'Ukrainisation' campaign against Armenia.