A newly introduced free, open-source tool called Latch is designed to stop AI agents from spending money unless a user explicitly approves the exact offer in front of them.
Latch · New open-source tool for AI agent payments
No AI agent spends money until the user approves the exact offer
Latch is a free control layer that sits between an agent and a purchase. Each quote carries a SHA-256 fingerprint of its terms, and the agent cannot spend until the user approves that specific quote.
5
steps in the described workflow
SHA-256
exact cryptographic fingerprint attached to each offer
0
spending allowed before the user approves that quote
How much of the five-step sequence is detailed
Each block is one step. The announcement's full sequence was not available in the material reviewed.
The core approval flow
STEP A
Agent requests a quote
→
STEP B
Offer carries a SHA-256 digest
→
STEP C
User approves that exact quote
→
Remaining steps of the five-step workflow are not detailed in the material reviewed.
Why it matters
Approval is tied to a fixed, hash-verified set of terms rather than a general instruction.
Transactions become easier to audit against the exact offer the user accepted.
Open source lets developers and security reviewers inspect the code.
What remains unverified
No independent testing, adoption figures, or payment provider statements yet.
Coverage across merchants and payment systems is unconfirmed.
How approval reaches users, and how quotes are checked against merchant systems, is unclear.
Handling of expired or altered off Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...