AI agents often run with the employee credentials that created them
Original title: AI agents inherit employee access, and nobody’s watching how they use it
Nudge Security telemetry shows a median of seven agents, with OAuth grants to OpenAI and Anthropic bypassing network controls.
Nudge Security’s AI Adoption in Practice 2026 report says employee-built agents in trackers, CRMs and automation tools often run on the creator’s OAuth grant or API key. Endpoint and identity tools then miss them, and logs attribute agent actions to the employee. OAuth connections to OpenAI and Anthropic, seen in most monitored organizations, move data cloud-to-cloud outside DLP. The company urges teams to inventory agents, review grants and rotate secrets pasted into prompts.
Unconfirmed: Figures come from Nudge Security customer telemetry, not a complete census of organizations.
- Among organizations with agents, 44% had more than 10.
- OpenAI appeared in 96.9% of observed organizations and Anthropic in 94.5%.
- Pasted content or uploaded files appeared in 18% of prompt activity.
- Nudge advises rotating any key or token that lands in a prompt.
- The report recalls the 2025 Salesloft Drift OAuth token theft.
From the original article
4 min read at helpnetsecurity.com
AI agents that employees build inside project trackers, CRMs and automation tools often run on the credentials of the person who created them and carry that person’s access to company data, according to Nudge Security. Source: Nudge Security The AI Adoption in Practice 2026 report draws on anonymized, aggregated telemetry from Nudge Security customer environments.
By Sinisa Markovic · 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.