The Watch

CISA added an actively exploited Fortinet FortiMail zero-day to its Known Exploited Vulnerabilities catalog Thursday, per The Hacker News; the flaw is an unauthenticated arbitrary file write on the appliance that terminates your inbound SMTP. Patch status alone won't tell you whether you were hit — the gateway holds DKIM signing keys, TLS private keys, and the LDAP bind credential used for recipient verification. An attacker who owns it reads mail in transit and signs outbound phishing that passes DMARC as your domain, so rotate those secrets, not just the build.

The Iran conflict now carries a stated date: President Trump said strikes resume after the November midterms absent a deal, with about 9,000 US troops and a third carrier heading to the region and an Iran-linked plot against a US-operated UK base just foiled, per Morning Brew. Iranian clusters such as Pioneer Kitten historically answer kinetic pressure by exploiting internet-facing VPNs and firewalls — the exact appliances already under a fresh wave of zero-days. Google Threat Intelligence Group separately found in-the-wild exploitation nearly doubled from January 2025 to August 2026, so your pre-November edge-hardening sprint is now dual-purpose.

Anthropic's new Claude Code mods are TypeScript add-ons installed via /plugin, distributed through a public directory, that run with the same machine access as Claude Code itself — and Claude will write one on request, per Simplifying AI. On the laptops holding your source code, cloud credentials, and signing keys, that is an unsandboxed extension supply chain with no vetting layer. Outflank separately showed Codex and Claude Code drift from the base MCP spec, so the schema the model sees can differ from what the server sent, weakening the human-in-the-loop control your policy assumes.

AI holding company Long Lake closed its roughly $6.3B take-private of Amex Global Business Travel at $9.50/share, moving your executives' itineraries and passport data to a new owner and off public-market disclosure, per Augment. In the other direction, Anthropic is targeting a November IPO — a draft prospectus obtained by Reuters shows ~$518B in compute commitments against $4.6B of 2025 revenue — which would subject it to SEC 8-K incident disclosure for the first time. For private vendors, your contract, not an SEC filing, is now the only thing guaranteeing you hear about a breach.