Feedzai launches AI Agent Detection aimed at automated fraud and MFA bypass

Feedzai has launched a capability called AI Agent Detection, aimed at automated financial crime, including attacks that get around multi-factor authentication, FF News reported on 13 November.

The product targets a shift that fraud teams have been bracing for. Older detection stacks were built to separate humans from scripts, and their signals lean on the assumption that automation looks different from a person. Agents weaken that assumption: they type, pause, navigate and retry much the way a customer would.

Naming MFA bypass as the threat is the sharper part of the announcement. If one-time codes and push approvals can be coaxed or relayed by an automated flow, account takeover gets cheaper for the attacker and harder to spot for the defender.

The available detail stops at the product's name, the threat it addresses and the fact of the launch. There are no published numbers on detection signals, false positive rates, coverage across channels, reference customers or pricing, so no claim about effectiveness can be made here.

Feedzai's move fits a broader pattern of defenders using AI to keep pace with attackers. As agent tooling has become cheap and accessible, the cost of automating the early stages of an attack has fallen, and vendors on the defence side are responding by treating agent behaviour as its own detectable category.

Three things are worth watching: whether detection works without adding friction for ordinary customers; whether banks and payment processors start writing agent-traffic identification into procurement requirements; and whether this layer complements or duplicates the device fingerprinting and behavioural biometrics already in place.

The harder question for financial institutions is not whether to defend against agents. It is where to draw the line when a growing share of legitimate agents act for real customers, and every block risks being a false alarm on a paying user.