RC RANDOM CHAOS

Scanner Impersonates 32 AI Bots to Hunt for Exposed Credentials

· via Dark Reading

Original source

Attacker Relentlessly Scans for Credentials by Impersonating AI Bots

Dark Reading →

A threat actor deployed a scanner mimicking 32 AI bots from OpenAI, Anthropic, Perplexity, Meta, and others to systematically search for exposed credentials across organizations in retail, financial services, telecommunications, energy, and software sectors. The scanner, active since late September and still operating through late October, fires bursts of 50 to 190 requests every 8 seconds while seeking passwords, API keys, cloud credentials, and SSH keys from approximately 90 file paths. On a single day in October, the malicious traffic hit 48% of organizations protected by Cequence Security’s monitoring.

The scanner revealed itself through behavioral quirks: rotating evenly through bot identities rather than favoring certain ones like legitimate bots do, claiming to be multiple vendors within seconds, and reusing Anthropic’s crawler contact address for all fake Claude identities. Nearly all requests were blocked or failed. However, security experts warn that while this credential harvesting largely failed, the broader threat extends to rogue agents logging into financial institution customer accounts using stolen or customer-provided credentials—a capability that concerns analysts more than the scanning itself.

Mitigation requires treating AI agents with the same rigor applied to human administrators: verifying agent identities through signed requests, published address ranges, or direct integrations; monitoring for behavior mismatches with claimed identities; and eliminating static credentials from vulnerable files in favor of centralized zero-knowledge secrets vaults.

Read the full article

Continue reading at Dark Reading →

This is an AI-generated summary. Read the original for the full story.