This science-for-policy brief presents findings from a 45-day observational Instagram field study conducted between 10 March and 21 April 2026 as part of the CARA Initiative. The study examines whether VLOP recommender systems can expose newly created, passive accounts to high-intensity content within an identity-sensitive masculinity ecosystem, without search, follows, posting, likes, comments or active engagement. The brief introduces CARA — Cohesion & Amplification Risk Assessment — as an operational measurement layer for observing relational risk, exposure trajectories, content intensity and algorithmic positioning in recommender systems. It summarises key field findings, including early Level 4 exposure in two new accounts, sustained Level 3 exposure in a neutral-entry account, and a calibrated CIS result in which Gate F was not activated. The document is intended for policy, regulatory and institutional audiences, particularly in relation to DSA systemic risk assessment, external audit methodology, ECAT/JRC-oriented discussion, and data-access needs under DSA Article 40. It is a policy-facing summary and differs from the full working paper in scope, format, length and intended audience.