Bay AI Launches With $50M to Vet AI Agent Actions On-Device

Bay AI Inc. launched today with $50 million in funding to intercept every action an AI agent proposes on a user's device before it runs, SiliconANGLE reported. The Israeli startup's argument is that first-generation AI security tools filter prompts and responses but miss where damage actually lands, since "a harmless-looking prompt can still end with an agent deleting a production database."

The product has three pieces. A scanner called Wave takes inventory of every agent, browser extension and local model on a machine. Bay Enforcer, which the company describes as a static binary, receives each proposed action through a native hook and returns verdicts in milliseconds using full session context. A third system named Lighthouse watches 30 days of agent activity and drafts separate policies per team, with Bay claiming false positives run under 1%. The install goes in through existing mobile device management or endpoint detection and response tools in under five minutes.

The $50 million breaks into a $37 million Series A led by Battery Ventures LP and a previously undisclosed $13 million seed co-led by YL Ventures and Striker Venture Partners, per a corroborating Globes writeup. Micron Ventures came in after Micron Technology had already deployed the product as a customer. Named customers include Wix.com, ClickUp and Peloton Interactive alongside Micron.

Co-founder and CEO Yuval Zacharia, CPO Aner Morag and VP of Research Boris Danilovich all worked together at Israel Defense Forces' Unit 8200; Danilovich then spent six years at CrowdStrike after it acquired Preempt Security. The launch lands in the same week we tracked OpenAI's internal control room for halting research models on command, one of a steady run of agent-safety pieces moving through our feed; different layer of the stack, same instinct that the intervention point has to move closer to where the agent runs.