Day's Snapshot by WRF – 11 Nov 2026
Five stories from the last day: SAP's four critical patches, the Fireworks AI breach, Google Cloud spend caps, N-able's outage, and insurance for AI agents.
SAP patches four maximum-severity flaws, none needing a login
SAP's November Security Patch Day fixed four flaws rated CVSS 10.0, all exploitable without authentication. Three sit in SAP NetWeaver AS ABAP components (ICM, the SAML 2.0 service provider and Gateway) and one in SAP Web Dispatcher. Onapsis, which found three of them, says it has not seen them exploited; advice includes patching every Web Dispatcher, restricting the Gateway proxy list and applying the SAML fix, which has no workaround.
Why it matters: If a partner runs your SAP landscape, ask for the patch date for each of these notes and who owns Web Dispatcher.
Fireworks AI breach widens to customer data and model weights
Fireworks AI says an unauthorised party reached its development environment through a third-party cloud provider on 23 October and ran a workload the next day. Exposed data includes billing information and some incomplete model weights and datasets for certain customers, but not payment card numbers. Notion says fewer than 1% of its workspaces were affected and has disabled its Insights feature; customers are advised to rotate credentials pasted during the window.
Why it matters: Ask AI partners which third-party inference providers they use and how credentials and data are isolated.
Google Cloud spend caps reach general availability
Google Cloud's spend caps block new requests to a capped service in a project once estimated spend hits the target, while in-flight requests finish and nothing is deleted. Caps reset monthly and now have an API. Enforcement covers only the Gemini API, Gemini Enterprise Agent Platform, Cloud Run and Cloud Run functions, and Provisioned Throughput for AI models is never halted.
Why it matters: A cap is a safety net, not a budget. Ask cloud partners what it does and does not cover in your setup.
N-able takes its cloud remote-management service offline
N-able took its cloud-hosted N-Central service offline and urged customers to take down self-hosted deployments after a researcher published reported zero-day bugs on GitHub on Sunday. The researcher says they can steal credentials, create admin accounts and gain root, and released them over unpatched flaws he had reported in another product. N-able says it has no evidence of successful exploitation.
Why it matters: If a managed IT partner uses N-Central, ask whether it is affected and what the fallback is for monitoring and patching.
Cowbell launches insurance for losses caused by AI agents
Cowbell Cyber unveiled a standalone policy for US small and midsized businesses covering losses caused by AI agents they deploy, including model drift, agent manipulation and LLMjacking. It covers first-party costs such as investigation and unrecovered funds, and third-party claims. It goes on sale on 1 January 2027 through wholesale brokers, and underwriting looks at observability, autonomy and governance.
Why it matters: Insurers are asking how agents are monitored and limited. Ask your AI partner to document that, not only the features.