Gartner, as cited by JetStream Security: 40% of enterprises will demote or decommission AI agents by 2027

Gartner, as cited by JetStream Security, says 40% of enterprises will demote or decommission autonomous AI agents by 2027 because governance failures are only catching them after production incidents (ACCESS Newswire). The company says the trust gap is the market it intends to fill.

JetStream is one of a new crop of vendors selling "agent governance": the rules, oversight, and stop-buttons that sit between an autonomous AI system and the business it acts on. The San Francisco startup said Monday it has been accepted into NVIDIA Inception, the chipmaker's startup program, and claims it is one of the fastest companies in the program's ten-year history to complete onboarding.

The company also cites FedRAMP High (Class D) certification and a Verified MCP Governance Layer launched earlier this year. Both are company-asserted: FedRAMP is a federal procurement posture, not runtime agent safety, and "verified governance" claims in this category are not yet independently auditable. No standards body has converged on what agent governance actually means.

JetStream previously raised a $34M seed round from cybersecurity investors, per independent reporting (Yahoo Finance). CEO Raj Rajamani frames "the recent OpenAI and Hugging Face security incident" as proof that trust is the industry's central question: one vendor's read on a real problem, in a market the vendors themselves are still defining.