Bitforest detects AI-agent attacks on Japanese company websites

Bitforest says it deliberately declined to explain exactly how it spots AI-agent attacks, warning that publishing the method would let attackers adapt. Instead it leaned on the very trait that defines these agents - reading the server's response and revising their next move - to separate them from older, simpler traffic.

The company says its pre-deployment testing caught attacks from eight high-performance models, but it also cautions that it cannot enumerate every AI agent. Its claim is narrower and stronger: if the system detects one, it is definitively an AI agent.

The three attacks Bitforest saw differed enough that they may have come from separate actors, which suggests this is not a single campaign. Bitforest's own conclusion is that AI agents are too useful for attackers to ignore, and it expects essentially every attacker to use them going forward.