Google just let the security industry know how it views the AI agent problem. It skipped the whitepaper and grouped a cohort instead.
The inaugural Gemini Startup Forum: Cybersecurity grouped 33 startups across six focus areas. Onyx is one of eight total vendors in AI agent security and governance, standing on its own as a full segment rather than a subset of application security or a tier of identity. Twelve months ago this specific grouping did not yet exist in any industry-standard analysis.
What this really means is that the hyperscaler that runs Vertex AI Agent Engine, Gemini Enterprise, and Google Workspace Studio is telling the industry that autonomous agents need their own security discipline.
The AI agent security cohort
Why this matters right now
The AI program at most enterprises is out-pacing their security team's capability for control. Leadership is looking for efficiency and return on AI spend while employees are quickly and eagerly adopting AI tools into their workflows. Legacy tools lack the necessary visibility and controls to properly secure AI adoption without hurting velocity.
Vertex AI Agent Engine, Gemini Enterprise, and Google Workspace Studio are all covered by Onyx today. The Startup Forum brings Onyx's product team into direct working sessions with the DeepMind and Google Cloud engineers building the next set of agent surfaces, which means enterprises standardizing on Google's AI stack get security coverage that keeps pace with what Google ships rather than lagging a release cycle behind.
Google picked eight companies to solve this problem and Onyx is one of them, so if your security team is looking to solve AI control, you should take a look.
If you want to see how Onyx handles Vertex, Gemini Enterprise, and Workspace Studio in a real environment, book a demo.
Frequently Asked Questions
The inaugural Gemini Startup Forum: Cybersecurity is a Google program that grouped 33 startups across six focus areas. It gives selected companies direct working sessions with the DeepMind and Google Cloud engineers building new agent surfaces, rather than publishing a whitepaper defining the market from the outside.
Google positioned AI agent security and governance as a full segment rather than a subset of application security or a tier of identity. Twelve months ago the grouping did not exist in any industry-standard analysis. The signal is that the hyperscaler running Vertex AI Agent Engine considers autonomous agents a separate discipline.
Eight companies were named: Capsule Security, Evoke Security, Manifold Security, Mirror Security, Onyx Security, Refractal, Unbound Security, and XOR. Approaches range from runtime protection and agentless detection to homomorphic encryption, cryptographic audit logging, and reinforcement-learning environments for finding security flaws.
Onyx covers Vertex AI Agent Engine, Gemini Enterprise, and Google Workspace Studio. Participation in the Startup Forum puts the Onyx product team in direct sessions with the engineers building Google's next agent surfaces, so enterprises standardizing on Google's AI stack get coverage that tracks what Google ships.
AI programs at most enterprises are outpacing the security team's capacity for control. Leadership wants return on AI spend while employees adopt AI tools into daily workflows. Legacy tooling lacks the visibility and controls to secure that adoption without slowing velocity, which is the gap agent-specific platforms are built to close.
