Why a physician is testing your AI agent.

The question everyone who Googles us asks. Answered straight.

The principal

Counteragent Security is led by Dr Sharmadave Subramaniam — a practicing physician who has spent years breaking systems for a living under the security-research handle TheDocter. Medicine and adversarial testing share a discipline: you form a hypothesis about how a system fails, you test it against reality, and you don't get to round up.

AI agents fail the same way patients do — quietly, at the boundary between instructions and trust. Our job is to find that failure before your buyer's security team does, and to hand you the evidence that closes the deal.

How we work

Every engagement is performed personally by the principal, assisted by a proprietary testing harness. No subcontractors without your written consent. Assessments are fixed-scope and fixed-price: scoping first, two-week assessment, retest included.

Evidence over claims

We publish our work. Our sample report — a full adversarial assessment of a reference agent, with measured attack rates and disclosed measurement corrections — is public, along with the reference agent and hashed run logs, so any reviewer can reproduce the results before trusting the method.

Read sample report ASA-2026-001 · Read the methodology