Bay Raises $50 Million to Control What Enterprise AI Agents Can Do

Bay, a new company founded in February 2026 by Yuval Zacharia, Boris Danilovich, and Aner Morag—all alumni of Israeli Military Intelligence Unit 8200—has secured $50 million in funding. This capital comes from a $37 million Series A led by Battery Ventures and an initial $13 million seed round from YL Ventures and Striker Venture Partners. The company aims to address the lack of visibility into autonomous AI agents within enterprises, which operate unpredictably at runtime based on user permissions. Bay functions as a central control plane that inventories, maps connections between, and monitors all deployed AI agents, assistants, and local models across an organization's environment.
The core function is the Enforcer component, which evaluates every proposed agent action in milliseconds against policies automatically derived from the organization’s historical activity. This design explicitly guards against dangerous activity stemming from either human-directed prompt injection or unintended "agent purpose drift." The founders emphasized that their development was guided by a focus on effortless deployment, rapid demonstrable value (achievable in under 14 days), and clear operationalization through AI-driven policy creation. Future plans involve establishing Bay as the fundamental trust layer for large-scale enterprise adoption of autonomous agents across the US and Europe.

Bay was founded in February 2026 by Yuval Zacharia, Boris Danilovich, and Aner Morag.
All three founders are alumni of Israeli Military Intelligence (IDF Unit 8200).
The company raised $50 million in funding.
This funding included a $37 million Series A led by Battery Ventures.
An undisclosed seed funding of $13 million was provided by YL Ventures and Striker Venture Partners.
Bay provides a control plane to discover, understand, and control AI agents.
The Enforcer component evaluates proposed actions based on session context in milliseconds.
Policy guidance for Enforcer is automatically built and maintained by Bay’s embedded AI system, Lighthouse.
The company states its goal is to be the core control layer for enterprises using AI agents at scale.
Development plans include expanding engineering staff and executing go-to-market strategies in the US and Europe.

Bay proposes a critical architectural shift: moving trust from pre-deployment validation to runtime governance of autonomous actions. The strongest narrative is that Bay solves the acute governance gap emerging as AI agents become more capable; it offers necessary structural oversight without stifling innovation, claiming to manage risks like prompt injection and purpose drift simultaneously. This aligns with an increasing industry anxiety about the opacity inherent in distributed agentic behavior—a significant societal friction point between rapid technological deployment and established risk management frameworks.
The underlying assumption driving this market narrative is that human-defined policy can successfully constrain emergent algorithmic agency across complex, dynamic environments. The claim of deriving policies "automatically... based on the organization’s own activity history" suggests a move toward adaptive governance, rather than static rule sets. However, this places immense trust in the system's ability to accurately capture and model an entire enterprise workflow without developing systemic biases or security blind spots derived from its training data.
The pattern here is one of 'Risk Commodification,' where inherent technological risk (agentic unpredictability) is packaged as a solvable engineering problem requiring bespoke oversight middleware. The beneficiaries are enterprises moving AI into high-stakes workflows, while the cost borne by users remains an abstract reliance on the efficacy of Bay's claimed enforcement mechanisms.
Patterns detected: none
Autonomy check: No autonomy signals.