AI agents that train and measure analytic judgment against hostile influence
Mindkeel is an AI defence-tech company in formation. Our platform uses AI agents to train and measure the judgment of intelligence and security analysts, in exercises and in daily work.
Mission
Defending people and institutions against adversarial influence operations
We help intelligence and security analysts recognise hostile manipulation and keep sound judgment under pressure. Training leads see what holds, what fades and when to reinforce it.
The challenge
AI makes influence cheaper to run; AI agents now make realistic defensive training possible at scale.
Influence targets the people who decide
Analysts are the first filter for hostile narratives, so their judgment is itself a target, alongside the public's.
Analytic training is a priority
After 7 October, the IDF Chief of the General Staff named critical thinking “a central pillar” of intelligence research (December 2025).
AI lowers the cost of influence
AI providers document state-linked actors using generative models to run influence operations more cheaply (OpenAI, 2024).
Resilience needs a measure
Training creates value when it shows how judgment holds months later, under pressure and in real work.
The platform: AI agents at the core
An AI-native platform for training and for daily work.
Multi-agent simulation
Swarms of AI agents play adversaries, media and synthetic populations in a closed range, building realistic, labelled exercises on documented manipulation techniques (DISARM), adapted to the cultural and regional context of each theatre.
Analysis agents
AI agents detect coordinated campaigns in open sources, so real threats become new exercises.
Calibration engine
Analysts assess sources and record probabilities and confidence. The engine scores their judgment (Brier score) before an exercise, right after it, months later and on real assessments.
Judgment dataset
Every cohort adds structured analytic judgments. Aggregated with customer consent, they improve scenarios and measures over time.
Published evidence: less than an hour of training improved forecasting accuracy by 6–11% (Chang et al., 2016). Mindkeel trains continuously and measures in daily work.
Built on published science
Each building block rests on peer-reviewed work. Our job is the engineering that joins them.
Prebunking builds resistance
Short inoculation videos improved recognition of misleading rhetorical techniques, in the lab and in a field test on YouTube.
Roozenbeek et al., Science Advances, 2022
Analytic judgment can be scored
Brief training improved the accuracy of probabilistic forecasts in four years of US intelligence-sponsored tournaments.
Mandel & Barnes, PNAS, 2014; Chang et al., JDM, 2016
Readiness needs several tasks
In 70 intelligence analysts, calibration gains on two tasks moved independently, so readiness is measured on several tasks.
Kelly & Mandel, Applied Cognitive Psychology, 2024
Coordinated campaigns leave traces
Coordinated networks can be surfaced from open social-media data with behavioural-trace methods.
Pacheco et al., ICWSM, 2021
Secure delivery
The platform runs on the customer's premises, with a local model and full audit logs.
On the customer's premises
Data and models stay inside the customer's environment.
Defensive by charter
We optimise defences. Exercises use synthetic, clearly labelled material and stay inside closed training ranges.
Consent and cohorts
Participation is voluntary and informed; results are reported by cohort.
Audited and vetted
Every exercise is logged and every customer is vetted under export-control rules.
Team
Massimiliano Gattoni
Co-founder, CEO & CTO
Computer engineer with more than 25 years of experience, over 20 at C-level, in digital, innovation and strategy. Former Enterprise Architect at Microsoft. Led digital and security for Italy's critical national infrastructure as CIDO of Open Fiber, Chief Technical Secretary to the Minister of Infrastructure and Transport, and CIO of ANAS. Founder & CEO of NeuroMind AGI; author of Cognitive Warfare (2026); Knight of the Order of Merit of the Italian Republic.
Roie Yellinek
Co-founder & CSO
CEO of Geopolitics Lab (G-Lab), a geopolitical consultancy. Geopolitical advisor to the Israeli National AI Directorate at the Prime Minister's Office. Formerly external strategic consultant and researcher for the Israeli Ministry of Defense (2021–2025), associate researcher at the BESA Center for Strategic Studies, adjunct researcher at the IDF Dado Center and non-resident scholar at the Middle East Institute in Washington. Writes extensively on geopolitics and soft power. PhD, Bar-Ilan University.
Contact
We welcome conversations with defence institutions, training commands and partners.
Write to usinfo@mindkeelai.com