Bynario, seven-person Milan security startup, raises EUR2.1m pre-seed led by 360 Capital

Seven people sharing EUR2.1m is about EUR300,000 a head ^[1][2][1], which pays for research salaries and not much else, and the ratio against the capital standing behind it is starker: 360 Capital manages more than EUR700m ^[10], the platform PranaVentures formed by merging with P101 SGR oversees more than EUR600m ^[11], and the whole round is roughly 0.16% of the EUR1.3bn between them ^[2]. On those round numbers 360 has averaged something near EUR9m per company across its 80-plus holdings ^[10][5], one of which is Equixly, another Italian cybersecurity startup ^[10]. This is an option written on three founders and a method, small enough that a wrong bet would barely register in a fund report.

The pitch is subtraction. Most scanners flag every theoretical weakness; Atlas is built to confirm which flaws are actually reachable in a given customer environment before prioritising and helping close them ^[5], so what is being sold is partly a promise about what the customer will not be sent. Pesoli's own argument is that cheap AI lowers the cost of vulnerability research ^[17], and it lowers it symmetrically for the person fabricating findings and the person finding real ones, which is how a review queue ends up full of what researchers call AI slop ^[6]. A vendor whose design goal is fewer, provable reports is selling reviewer hours back.

Escape's $18m in March, Qevlar AI's $30m the same month and Mindgard's $30m in August come to $78m ^{[13][14][15][3]} of disclosed capital aimed at neighbouring problems, and by techfundingnews's reading none of the three runs the whole loop from discovery through proof of exploitability to remediation ^[18]. That gap is either a moat or a line item on a larger vendor's roadmap, and EUR2.1m does not buy the distribution that settles which. 360 partner Cesare Maifredi frames the requirement as autonomous discovery, validation, prioritisation and remediation ^[16], which is also a description of four separable products, any one of which an existing scanner can bolt on and sell to accounts it already has.

The weakest leg is the breach-vector story. Techfundingnews puts exploited vulnerabilities at 31% of breaches, ahead of stolen credentials, and names no study behind the figure ^[12]; on that same number, 69% of breaches begin with something other than an exploit ^[4], so the leader of this league table holds under a third of the field. What the evidence does carry is narrower and more useful: a company rationing its review queue by quota is a buyer putting a price on noise ^[6], and at this cheque size Bynario only has to be right about that price.