Arrakis: 76% of AI agents sat idle 60 days with credentials still live
Arrakis Security telemetry from thousands of production agents found 76% had not run in 60 days while their credentials still worked, and 82% carried at least one exposure.
Arrakis Security released its 2026 Autonomous Agents Threat Report this week, based on anonymized production telemetry from thousands of AI agents in its customers' environments, collected over 12 weeks1. The headline: 76% of active agents had not run in 60 days, yet their credentials still worked and their triggers still fired, and 44% had not run in 180 days1. Arrakis also found that 24% of dormant agents could be started by an unsigned webhook from outside the organization, and that 82% of agents carried at least one exposure, with 46% of those carrying more than one1.
Help Net Security's coverage adds detail on what those idle agents can reach2. Seventy percent of agents connected to at least one integration classified as sensitive, and 99.4% of those had no verified data loss prevention coverage on that connection2. Nineteen percent used API keys and tokens that never expired and were embedded in their configurations2. Tal Baron, Arrakis CEO, put the cause plainly: when an agent stops being used, nobody removes anything, because in most companies nobody owns it2.
This is vendor data from a security company that sells agent posture tools, and the sample is its own customers, so treat the percentages as a signal rather than an industry baseline. Still, the pattern is familiar. Teams are quick to grant an agent access to ship a pilot and slow to take it back when the pilot ends. An idle agent with live keys is an unused service account with a trigger attached.
For data teams, the warehouse is often where that access lands. An agent's database role, its stored credentials and its scheduled triggers should all have a last used date we can query. Access logs and query history already show which identities have gone quiet. Long lived keys belong in a secrets manager with expiry, and lineage should show which tables a dormant agent could still read or write if someone woke it up.
One step this week: pull the last login or last query time for every non human identity with warehouse access. Any agent idle for 60 days gets its grants revoked and its keys rotated, and the owner can ask to restore access if it still has a job.
- Arrakis Security via GlobeNewswire (CIO Influence) · Oct 22, 2026
Arrakis Security Report Finds 76% of AI Agents Are Dormant but Keep Their Credentials and Connections - Help Net Security · Oct 26, 2026
Forgotten AI agents may still access sensitive company data