Anthropic tells customers a third-party cybersecurity firm accessed internal systems without approval on Oct 8, possibly exposing Claude business contact and usage data

Anthropic customer notice reported by Byline Network (Nov 7) and customer-shared email on X; Anthropic has not posted a public English statement; firm unnamed; likely related to the pending Cathedral authorized-test dispute2026-11-07

Anthropic tells customers a third-party cybersecurity firm accessed internal systems without approval on Oct 8, possibly exposing Claude business contact and usage data

Not in scope (per notice): Claude prompts/replies, passwords, credit cards, bank accounts.
Do now: treat unexpected Anthropic emails or calls about billing, plan changes or "security verification" as phishing until confirmed in-product; do not send passwords, OTP codes or payment details.
Likely related to the Oct Cathedral AI-agent access story (still pending a primary Cathedral write-up); the notice does not name the firm.

Sources: Byline Network: Anthropic unauthorized internal access / Claude account info risk (Nov 7, Korean) · Jamieson O'Reilly customer notice post on X (Nov 7)