Critical Vulnerabilities Patched in Arista, GitLab, and Keycloak; Banks Detect AI Agents

Critical Patches Released

Arista patched a critical SQL injection vulnerability (CVE-2026-114625, CVSS 9.9) in its on-prem VeloCloud Orchestrator. This flaw allowed any authenticated user to gain operator privileges and impersonate a 'superuser operator'. Hosted VCO instances are already patched, while on-prem administrators need to upgrade.

GitLab released a critical patch on October 21, 2026, for CVE-2026-110264, an integer overflow in the bundled Onigmo / regex compiler. This vulnerability, rated CVSS 10.0, could allow an unauthenticated attacker to execute code on self-managed GitLab instances. All CE/EE versions from 7.0 (2014) are affected, and instances that only applied a previous regex fix must upgrade again.

Keycloak addressed a flaw in its reset-credentials flow that permitted an unauthenticated attacker to force a password reset for any user without verification, leading to account takeover. Red Hat rated this vulnerability CVSS 9.1. Fixes were shipped in August (26.4.15, 26.6.6, 26.7.2), and a full proof-of-concept was published on Exploit-DB on October 21, urging immediate upgrades for unpatched instances.

AI Agents in Banking Accounts

Axios reported that major U.S. banks, including Morgan Stanley, Bank of America, JPMorgan, and Wells Fargo, are detecting third-party AI agents operating within customer accounts. These agents often log in using the customer's own credentials, which violates bank rules requiring individuals to log in themselves.

Some banks also observed agents contacting call centers or performing actions that appeared independent of user instructions. Bank executives noted difficulty in distinguishing between human users and AI agents, or between authorized and rogue agents. No account takeovers, fraud losses, or specific AI agent products have been reported. Morgan Stanley is blocking these agents, while JPMorgan, Bank of America, and Citi are restricting them.