Arrakis says dormant AI agents keep access to company systems
Original title: Forgotten AI agents may still access sensitive company data
In a 12-week review, 76% of enabled agents had not run in 60 days and 22.9% had an unsecured trigger.
Arrakis Security’s 2026 Autonomous Agents Threat Report says inactive AI agents in customer environments still hold credentials and connections. The analysis covered thousands of production agents over 12 weeks. About 76% of enabled agents had not run in 60 days, 44% had not run in 180 days, and about a quarter of dormant agents could still be started through an unsigned external webhook. Seventy percent connected to at least one sensitive integration, and 99.4% of those lacked verified data-loss-prevention coverage on that connection.
Unconfirmed: Findings come from Arrakis customer environments; the Hugging Face episode is a separate METR investigation, not part of the agent sample.
- Nineteen percent used API keys or tokens that never expired and were embedded in configurations.
- Eighty-two percent of agents had at least one recorded exposure; 46% of those had two or more.
- A cited METR note says about 700 of 1,200 agents later joined a Hugging Face attack.
- Some scheduled agents run weekly, monthly, or quarterly, so inactivity alone is not misuse.
Impact: Dormant agents can still read, write, and start workflows
- 1.Remove credentials and connections from agents that are no longer used.
- 2.Disable unsigned webhooks and other triggers an agent cannot verify.
- Vendors
- Arrakis Security
- Organizations
- METROpenAIHugging Face
From the original article
3 min read at helpnetsecurity.com
AI agents that have not run recently retain credentials and connections to company systems, according to Arrakis Security’s 2026 Autonomous Agents Threat Report. About 76% of enabled agents examined had not run in 60 days, leaving access available through inactive software.
By Anamarija Pogorelec · helpnetsecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.