Zero Radar
← feed
HighSecurity incidentupdated 28 Oct 2026, 14:55 UTC

Fake AI Bots Scanning Systems to Steal Exposed Credentials

Attackers are disguising their scanning programs as legitimate AI company bots. If your systems leak passwords or access keys, these bots can find them and let attackers in.

Attacks reported (Dark Reading)1 source, unconfirmed
SeverityHighSecurity incident
Being exploited?ReportedBy Dark Reading, not officially confirmed yet
What to doCheck exposureRotate any credentials that may be publicly exposed and remove them from accessible locations.
How sure is this?1 source, unconfirmedOnly one publisher so far, details may change

How the attack works simplified

  1. 1Threat actorOperates disguised scanning bots
  2. 2Impersonated AI botsClaim to be OpenAI, Anthropic, Meta agents
  3. 3Exposed systemsProbed for exposed data and secrets
  4. 4Credentials foundPasswords or keys harvested from exposure
  5. ResultAttacker gains usable credentials

What happened

A threat actor ran bots impersonating agents from OpenAI, Anthropic, Meta and others to search for exposed credentials and data.

Why it is dangerous: Organizations that expose credentials or sensitive data publicly, and that allow traffic from AI crawler-looking user agents, could have secrets harvested and used for further access.

Are you affected?

You are affected if your web services or code repositories expose credentials or sensitive data to automated traffic, including bots claiming to be AI agents.

What to do

  1. Remove any credentials, API keys or secrets exposed in public code, configs or web pages and rotate them
  2. Check server logs for traffic from bots claiming to be OpenAI, Anthropic, Meta or other AI agents and verify them against each vendor's published IP ranges
  3. Restrict access to sensitive endpoints and block unverified automated scanners

Timeline

  1. 28 Oct 2026, 14:37 UTC
    Publicly disclosed
  2. 28 Oct 2026, 14:37 UTC
    Reported by Dark Reading · Dark Reading
  3. 28 Oct 2026, 14:55 UTC
    Picked up by Zero Radar

Sources 1 report, oldest first

Security press established security newsrooms

Technical details

ID
ZR-70a17c1d
Disclosed
28 Oct 2026, 14:37 UTC
On Zero Radar
28 Oct 2026, 14:55 UTC (18 min after disclosure)
Last change
28 Oct 2026, 14:55 UTC

The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works

Fake AI Bots Scanning Systems to Steal Exposed Credentials · Zero Radar