# COSIGN

> The authorization layer for the AI era. COSIGN halts every consequential action — outbound wires, vendor and beneficiary bank-detail changes, payroll edits, large refunds, and credential/access resets — until a named, authorized human (or policy-bound quorum) approves it with a cryptographic, out-of-band cosign. It does not try to detect deepfakes; it makes the legitimate path provable and refuses everything else. Every approval is sealed in an immutable, hash-chained ledger and backed by a written, insured loss guarantee.

## Core thesis
- The problem is not that fakes are hard to spot; it is that trust is anchored to faces and voices, which AI now forges trivially (a voice clone needs ~3 seconds of public audio; video deepfakes pass live calls).
- Detection is probabilistic and degrades every model cycle — an arms race the defender loses. COSIGN is deterministic: a hardware passkey + biometric signature, bound to the exact transaction, that a synthetic voice/video cannot produce.
- Friction is applied only to the risky path (over-threshold amounts, new beneficiaries, bank-detail changes). Routine in-policy payments flow normally.

## Why now (June 2026)
- The bipartisan Great American Artificial Intelligence Act — a June 2026 discussion draft, not yet enacted — proposes doubling federal penalties for AI-enabled wire, bank, and mail fraud, and would preempt state AI *development* law (with a sunset) while preserving state AI *use/deployment* law, increasing deployer liability.
- AI-enabled fraud surged ~1,210% (Pindrop, 2026); U.S. AI-fraud losses are projected at ~$40B by 2027 (Deloitte, from $12.3B in 2023). The FBI IC3 reported $20.9B in total 2025 losses, including $3.05B in business email compromise. Reference incident: engineering firm Arup lost $25.6M in a single deepfake video call (2024).
- Card networks (Visa Intelligent Commerce, Mastercard Agent Pay, Google AP2) are building agent->merchant *consumer* checkout — none address internal treasury/AP/payroll/identity money movement. That gap is COSIGN's market.

## How it works
1. INTERCEPT — connects to bank, ERP, AP, payroll, and identity provider; freezes consequential actions above configured thresholds.
2. COSIGN — pushes the exact transaction to a named approver's registered device for passkey + biometric approval, cryptographically bound to that transaction; optional 2-of-3 quorum.
3. LEDGER — seals who/what/when/which-device in an immutable hash-chained record for audit, compliance, and insurance evidence.

## Who it is for
Mid-market companies ($50M-$1B revenue) — Treasury & Cash Ops, Accounts Payable, Controllers/CFOs, Payroll, and IT/Identity teams that move money and credentials daily.

## Offer
- Free 30-minute Fraud Audit: maps every path money/credentials can leave the company, shows where a deepfake walks through, and prices the guarantee.
- The COSIGN Guarantee: if fraud clears a COSIGN-protected action, the loss is covered in writing, up to the tier's policy limit (underwritten via insurance partner; deterministic control makes it underwritable).

## Pricing
- Treasury Guard — $1,500/mo: single entity; wire + vendor bank-change protection; up to 5 approvers; $1M guarantee limit.
- Command — $4,500/mo: bank + ERP + payroll + identity; quorum approvals & custom thresholds; unlimited approvers, SSO, SOC 2; $10M guarantee limit; quarterly deepfake red-team.
- Enterprise — custom: multi-entity; AI agent-action cosign API; dedicated underwriting; on-prem ledger option.
- Unit economics: the average AI-fraud incident is $280K-$680K; Command pays for itself if it stops one attempt per decade.

## Expansion
The same rail authorizes AI-agent actions: it stops a prompt-injected or compromised agent from moving money on its own authority. As agents transact autonomously, every consequential agent action will require a verifiable, policy-bound cosign. Beachhead: human payment-fraud prevention. Category: the authorization standard for the agentic economy.

## Key pages
- Home: https://usecosign.com/
- How it works: https://usecosign.com/#how
- Guarantee: https://usecosign.com/#guarantee
- Pricing: https://usecosign.com/#pricing
- FAQ: https://usecosign.com/#faq
- Security: https://usecosign.com/security
- Agents: https://usecosign.com/agents
- Book a fraud audit: https://usecosign.com/#book

## Contact
- Book a fraud audit via https://usecosign.com/#book