A cluster of critical security disclosures and breaches landed within about 48 hours in late October, a stretch the AI community has dubbed 'Black October.' Among them were an unauthenticated remote code execution flaw in PHP, a CVSS 10 vulnerability in GitLab affecting releases dating back to 2014, and a Brotli decoder fix in Node.js. A release of the LiteLLM package, version 1.107.0, shipped with a password stealer. Railway told customers to rotate secrets after possible exposure of customer keys, and the Zeabur hosting service is shutting down after 321,810 accounts were breached. Certificate authority eMudhra's signing keys were reportedly so weak that one was the number 4. A researcher says Claude cracked them with a single prompt, and eMudhra is exiting the certificate authority business. Quest Diagnostics, which serves about half of US hospitals and physicians, also reported exposed patient data. Attribution and timing are contested. Microsoft says China-aligned group Storm-0940 used the open agentic penetration-testing framework ARTEX against more than 2,250 Microsoft 365 tenants, at about 2,900 sign-in attempts per minute, and in some cases reached privileged access in under three hours. CrowdStrike documented four operators running AI-agent campaigns, one of which hit at least 15 Argentine healthcare organizations. The PHP bugs were reportedly found by OpenAI models in August and held under embargo until a fix was released this week, and the PHP Foundation says Claude Mythos had scanned PHP earlier. Some commentators dispute the narrative that offense is winning. OpenAI's Joshua Achiam reportedly said only about four of the roughly 16 major incidents occurred this week, with the rest found earlier and disclosed now. On the defensive side, Anthropic's Frontier Red Team, working with Theorem Labs, used Claude to prove security properties across more than 100 open-source libraries. Anthropic's deputy CISO, Jason Clinton, warned that conditions 'are going to get a lot worse before they get better.' The story also reached financial services. Axios reports that major US banks are detecting unauthorized third-party agents inside customer accounts, some of which appear to act beyond user instructions. Other observers read these as user-authorized agents that the banks had not approved, rather than hacks. Senator Adam Schiff has called for mandatory industry-wide AI guardrails. For Australian businesses and professionals, the practical steps are concrete. Any team running LiteLLM, Railway, GitLab, PHP, or Node.js in production should check exposure now, rotate API keys and secrets, and update pip, uv, certifi, and system CA stores as the Python advisory recommends. Organizations holding personal information are covered by the Privacy Act 1988 and its Notifiable Data Breaches scheme, so a breach at a third-party vendor that exposes customer data could trigger notification duties to the OAIC and affected individuals. Teams that let third-party AI agents access customer accounts should review what those agents are authorized to do.