Skip to content

Response, New York. Noah Park: Check Point Patches Actively Exploited Management Server Path Traversal. Detection, Dallas. Ava Okello: Report Finds Nearly Half of Deployed Detections Need Attention Before SOCs Can Trust Them.

SOCtember

Always First. Fast SOC News.

Cequence CQ Prime Research chart titled The fake AI bot scanner's addresses lit up in mid-October, plotting 109 cloud addresses seen impersonating AI bots on October 21, 2026, with third-party blocklist flags and requests blocked by Cequence over the previous 30 days.
Graphic: CQ Prime Research, Cequence Security (fair use).

Detection

Cequence Says a Single Scanner Posing as 32 AI Bots Hunted for .env Files, Cloud Credentials and SSH Keys, Reaching 48% of Its Customers With AI Bot Traffic in One Day

The CQ Prime Research team says the actor rotated through crawler and assistant identities from OpenAI, Anthropic, Meta and Perplexity, burned through more than 300 cloud addresses in a day, and gave itself away by asking for secrets no real AI agent needs. Its advice is to verify agents by signed requests or published address ranges and to flag any claimed assistant that requests credential files.

Ava Okello, DetectionLondon5 min read

LONDON - A single scanner has spent the past month posing as roughly 32 different AI bots, including crawlers and assistants from OpenAI, Anthropic, Meta and Perplexity, while it combed websites for passwords, API keys and cloud credentials, according to research published Wednesday by CQ Prime Research, the threat research team at Cequence Security. On one day in October, the actor reached 48% of the Cequence-protected organizations that saw any AI bot traffic.

The finding lands as banks and other businesses decide which AI agents to let into their applications, and it undercuts the simplest way of making that call. "Attackers can test every one of those allow and deny decisions, and Cequence is seeing them probe which agents a business trusts, then impersonate the agents they expect to get through," wrote Will Glazier, director of CQ Prime threat research and data science. "For anyone responsible for application and API security, a bot's self-reported identity no longer proves anything."

What the scanner asked for

The scanner runs from a major cloud provider and ignores ordinary web pages. It requests the files where secrets tend to sit, including .env configuration files, cloud credential files, private SSH keys and server environment data, working from a list of about 90 paths. Each address fires a burst of 50 to 190 requests in about eight seconds and moves on, working through targets one at a time, around the clock. It opens with random nonsense paths to find sites that answer every request, a common scanning technique.

Cequence illustration of a bot disguising itself, used on the CQ Prime Research post about AI agent impersonation.
Illustration: Cequence Security (fair use).

Cequence says it hit organizations in financial services, retail, telecommunications, energy and software, and stayed active from at least late September through late October. Nearly all of its requests ended in a block or a failure, the company says, so what it documented is a sustained hunt rather than a confirmed breach.

Why the disguise gave itself away

For a detection team, the useful part of the research is how the impostor slipped. Real AI bots behave predictably. This one cycled evenly through all 32 identities, something no legitimate bot does. The same machine claimed to be OpenAI, Anthropic, Meta and Perplexity within seconds. All of its fake Claude identities carried the contact address of Anthropic's crawler, even when it posed as a different Claude agent, which Cequence reads as a copied and pasted string. And it never identified itself or offered a contact address, as legitimate internet scanning services do.

Screenshot of the CQ Prime Research section How Cequence Spotted the Fake AI Bots, listing the scanner's even rotation through 32 identities, multiple vendor claims from one machine within seconds, a copied Anthropic crawler contact address, and the absence of any self-identification.
Screenshot: CQ Prime Research, Cequence Security (fair use).The behaviors that separated the impostor from real AI agents.

Static blocking would not have kept up. The actor used more than 300 cloud addresses in one day and nearly 400 in one week, and none of the addresses from late September reappeared a month later. Third-party threat feeds flagged many of them on the same days they were used, Cequence says, but by then the scanner had usually moved on. What stayed constant was behavior: the same disguises, probe paths and request patterns at every target.

The impostor also distorts the metrics many teams now track. At several organizations in the study, it accounted for more than 80% of the traffic labeled as AI bot traffic. At the financial services firms CQ Prime studied, AI agents acting on behalf of customers made up fewer than 1 in 4,000 requests, even at the most exposed institution. "The agents banks worry about most log in with a customer's own credentials through an ordinary browser and don't announce themselves," Glazier wrote.

What to change

Cequence's guidance is to stop trusting the name a bot gives itself. Verify AI agents through signed requests, published address ranges or direct integrations. Watch for behavior that does not match the claimed identity, such as an AI assistant requesting configuration or credential files. Keep secrets out of web-accessible files, and confirm that staging and admin sites are not exposed to the internet. Evaluate bot traffic by how it behaves, and expect IP blocklists to go stale within days against this kind of activity.

Screenshot of the CQ Prime Research section Agentic AI Security Starts With Verifying Identity, listing recommendations to verify AI agents by signed requests or published address ranges, watch for mismatched behavior, keep secrets out of web-accessible files, and expect IP blocklists to go stale.
Screenshot: CQ Prime Research, Cequence Security (fair use).Cequence's recommendations for verifying AI agent identity.

In practice, that turns a vague worry about AI agents into a concrete web-log query: any request whose user agent claims to be a known AI crawler or assistant and asks for a .env file, a cloud credentials file or an SSH key is not an AI agent doing its job, and the source address deserves the same treatment as any other credential scanner.

Sources:


Ava Okello covers detection engineering, EDR telemetry, and SOC hunting for SOCtember from London.

Related stories

Detection desk