A new tally of vulnerabilities associated with Anthropic has reached 225, according to security firm Glasswing, but the practical impact so far appears minimal. The Register reports that VulnCheck's tracking of these CVEs has confirmed only one instance of exploitation in the wild.
The large gap between the number of disclosed flaws and actual attacks suggests that volume alone is not a reliable indicator of risk. Many vulnerabilities may require specific conditions or access that attackers have not found worthwhile. The source does not provide details on the nature of the flaws or why exploitation has been so rare.
Even so, the single confirmed case shows that at least one of these Anthropic-linked issues has been actively used. That keeps the set of vulnerabilities from being purely theoretical, and it may warrant attention from teams running affected systems.