AI · 6h ago
Reflectiz said its AI pentesting agent completed a live login bypass on a major insurer’s production website before its safety guardrails stopped the test. The agent was running an authorized assessment and was able to get a one-time passcode sent to an inbox it controlled, then log in successfully.
The flaw was in the login flow: the request let the user pick the delivery channel and destination for the one-time code, and the server sent the OTP there instead of to an address already bound to the account. In plain terms, the code still existed, but it no longer proved account ownership because the backend accepted a user-chosen mailbox.
For teams that let customers steer OTP email or SMS routing, the exposure sits in account recovery and authentication logic, not in the AI system that found it. If the destination is not locked to the account, a normal login can become account takeover once an attacker knows the identifiers the flow expects.
1 source covering this story
An AI Agent Found a Critical OTP Bypass in Production. The Attack Was the Easy Part.
Reflectiz's AI agent bypassed an insurer's live login by directing a one-time code to its own inbox during an authorized assessment.
Part of the PlainSec briefing for 2026-10-14