Fingerprint Verifies AI Agents, Claims Market Lead

Fingerprint, a US‑based device intelligence platform, launched Authorized AI Agent Detection to separate permitted AI agents from bad bots using Web Bot Auth - HTTP message signatures that let agents cryptographically prove who they are. Fingerprint says it now detects the highest number of AI agents on the market; for registered agents, verification is “with 100% certainty.” (thepaypers.com)

From blanket bans to trust‑by‑identity

• How it works: Agents sign requests with HTTP Message Signatures and a Signature-Agent header; sites verify against the agent’s public‑key directory per active IETF drafts. In short, identity over heuristics. (docs.fingerprint.com)

• Who’s in the starter pack:

• OpenAI ChatGPT Agent (OpenAI’s computer‑use/Operator agent). (openai.com)

• AWS Bedrock AgentCore (Amazon’s enterprise agent runtime). (aboutamazon.com)

• Browserbase (serverless browsers for agents and automation). (browserbase.com)

• Manus (autonomous agent startup acquired by Meta in Dec 2025). (apnews.com)

• Anchor Browser (cloud‑hosted browser built for AI agents). (anchorbrowser.io)

Why it matters

• Enterprises have been stuck between “block all bots” (break workflows) and “allow everything” (invite fraud). Trust‑by‑identity lets fintech and ecommerce allow selective transactions and automated purchasing while defending against ATO and payment fraud. (thepaypers.com)
• Fingerprint’s docs now list signed agents like chatgpt_agent, bedrock_agentcore, browserbase_agent, manus_agent, and anchor_browser in its Bot Directory, so customers can treat them as “good bots” in policy. (docs.fingerprint.com)

Availability and onboarding

• Available now and included for existing customers. New participants can contact the company and test signatures in the Web Bot Auth verification playground (a sandbox that validates agent HTTP signatures). (tmcnet.com)
• Implementation lives inside Fingerprint’s Bot Detection signal (Server API v4+); Pro Plus/Enterprise tiers apply. Translation: plugs into existing fraud stacks via the normal server‑side check. (docs.fingerprint.com)

Caveats, because security

• This rides on emerging standards; spoofing attempts will follow, and operational trust still hinges on how much weight you put on Fingerprint’s signal. Still, it aligns with broader industry moves to formalize agent identity-and observers expect agent‑driven transactions to surge by late 2026. (docs.fingerprint.com)

What’s missing

• Pricing wasn’t disclosed in the launch materials, and there’s limited third‑party validation beyond partner shout‑outs (e.g., Anchor). Net‑net: practical upgrade for fraud teams tired of false positives, but keep your threat models sharp. (tmcnet.com)