
Security for autonomous AI agents
TL;DR. Bay sells security controls for AI agents. An AI agent is software that takes actions on its own. Bay says it finds agents, checks their settings, and blocks risky actions. Its target users are enterprise security teams. Bay launched in November 2026 with $50 million in funding. We found no independent tests or audit reports.
Bay is a company that sells security controls for AI agents. An AI agent is software that takes actions on its own to complete a task. Globes calls Bay an "Israeli AI security platform" (Globes).
Yuval Zacharia, Aner Morag, and Boris Danilovich founded Bay in early 2026 (Globes). Bay announced itself publicly on November 10, 2026 (Globes).
The founders said that one of their coding agents tried to expose a cloud environment to the internet. This event occurred during a test. They said this event gave them the idea for the company. They also said no attacker controlled the agent. They said the agent used legitimate permissions (Calcalist).
Globes reported that Bay raised $50 million in funding (Globes). A seed round is a company's first funding round from investors. A Series A is the next early funding round. Calcalist reports that Battery Ventures led a $37 million Series A. Calcalist reports that YL Ventures and "Striker" led a previously undisclosed $13 million seed round (Calcalist). Micron Ventures, M12, and Assaf Rappaport participated in the Series A.
Jeetu Patel and Manoj Apte are also private investors (Calcalist).
On November 10, 2026, Globes reported that Bay had 35 employees. Most of these employees are in Israel (Globes).
The headquarters location is unknown. The launch press release was not available for review.
News reports name three co-founders (Calcalist). Yuval Zacharia is chief executive officer (CEO). Aner Morag is chief product officer (CPO). Boris Danilovich is vice president of research and development (Globes). The reviewed Bay pages did not confirm these titles.
Calcalist reports that Zacharia led a department of approximately 50 people during her military service. Later, she was director of research and AI at Hunters (Calcalist).
Calcalist reports that Morag joined Noname Security among its earliest employees. Later, he became VP of Product there (Calcalist). SiliconANGLE says Morag was the first employee at Noname Security (SiliconANGLE).
Calcalist reports that Danilovich led a department of approximately 90 engineers at CrowdStrike (Calcalist).
Calcalist reports approximately 35 employees, most in Israel (Calcalist). Globes reports 35 employees, most in Israel (Globes). Bay plans to expand its product and engineering teams. It also plans to grow sales, marketing, and customer teams in the United States and Europe (Calcalist).
Calcalist names Assaf Rappaport, Jeetu Patel, and Manoj Apte only as investors (Calcalist). We did not find board members or advisors in the evidence.
Bay's website lists three capabilities: "AI Asset Discovery", "AI Security Posture Management", and "Runtime Enforcement" (Bay). The website names laptops, browsers, cloud workspaces, and agents that work on their own (Bay). Bay says the time from deployment to automated prevention is less than 1 hour (Bay). This is a vendor claim.
Bay announced that Fortune 100 companies used its platform to triple their use of autonomous AI agents. The Next Web says Bay did not name those customers. Bay also did not say how it measured the figure. The Next Web did not verify the figure (The Next Web).
Other companies sell similar products. Zenity describes discovery, posture management, access control, runtime policy enforcement, and threat detection for AI agents (Zenity). A Series C is a later funding round. Zenity announced a $125 million Series C on August 3, 2026 (Zenity).
Straiker describes AI agent discovery, autonomous attack testing, and runtime defense (Straiker). Autonomous attack testing uses software to simulate attacks. Runtime defense means protection while agents run. Straiker announced a $64 million Series A on June 29, 2026. This round brought its total funding to $85 million (Straiker).
Recommendation: A buyer must ask Bay for measured results before purchase. We found no independent tests that compare Bay with competitors. We found no verified data on false positives or security effectiveness. A false positive occurs when a tool flags a safe action as risky.
A buyer can test policies in simulation mode (Bay). Simulation mode shows what a policy would do. It does not block actions.
Bay calls itself "the single control plane for all agentic activity, wherever it runs" (Bay). A control plane is one central place to see and control systems. Bay groups its capabilities as AI asset discovery, AI security posture management, and runtime enforcement (Bay).
Discovery: Bay says it finds every agent, extension, package, MCP server, skill, and model. Bay says it also classifies each of these items (Bay platform). An MCP server is a service that gives AI agents access to tools and data.
Bay says it maps the user behind each agent and the access of each agent. Bay also says it maps the MCP servers that each agent touches. Bay says it also maps the software as a service (SaaS) apps that each agent touches (Bay platform). SaaS apps are apps that run online.
Posture: Bay says it checks for signs of harm before approval. It checks agents, packages, extensions, MCP servers, and skills. Bay also says it finds unsafe settings, toxic combinations, and credentials exposed to MCP servers (Bay platform). Toxic combinations are tools or settings that are risky together.
Runtime enforcement: Bay says it reads each agent's whole session to understand intent. It then decides "allow, ask, deny or steer" on every action. Bay says it makes each decision on the device within milliseconds. Bay says it does not wait for a cloud service (Bay runtime).
Bay says Lighthouse, its policy-building system, learns from 30 days of past agent activity. Bay says Lighthouse also adjusts rules for each team (Bay runtime). Administrators can run any policy in simulation mode before enforcement (Bay platform).
Browser: Bay says it finds AI tools, extensions, and accounts in every browser. It ties every prompt and agent action to a corporate or personal identity. It can apply Allow, Warn, Steer, or Block controls for each team (Bay browser).
Cloud workspaces: Bay says it discovers agents across virtual machines, Kubernetes, Coder workspaces, and hosted sessions (Bay cloud workspaces). A virtual machine acts as a separate computer. Kubernetes is software that manages containers. A container holds an app and the files it needs to run. Coder workspaces are online development environments.
Deployment and integration: Bay advertises "Live in under 5 minutes" (Bay). We found no independent measurement.
MDM (mobile device management) tools manage company devices. EDR (endpoint detection and response) tools protect devices from attacks. AWS Marketplace says Bay deploys through existing MDM or EDR tools across Windows, macOS, and Linux. It also says Bay's MCP/API interface lets customers query findings, manage policies, and pull Bay data (AWS Marketplace). An application programming interface (API) lets software use functions in another system.
Ask Bay which tools your deployment can use.
Pricing on AWS Marketplace for 12 months: Bay Endpoint is $24,000.00 for 100 endpoints. Bay Cloud Workloads is $12,000.00 for 100 cloud workloads. Bay Browser is $8,000.00 for 100 browser licenses, as an add-on for Bay Endpoint (AWS Marketplace). Endpoint and Cloud Workloads bill independently in blocks of 100. Bay Browser requires Bay Endpoint (AWS Marketplace).
Bay's privacy policy names Bay AI Inc., Bay Security Ltd., and their affiliates as "Bay". The policy covers bay.io and the Bay platform at platform.bay.io (Bay privacy).
The policy says Bay collects full name, email address, and password. Bay also collects customer contact and billing information. It also collects information about actions taken through the Services and Website (Bay privacy). Bay uses personal information to provide the Services, process requests, and administer and improve the Services (Bay privacy).
Ask Bay how long it stores personal information.
Bay last updated the Cookies Notice on October 12, 2026. It says PostHog sets analytics cookies only after you accept the cookie banner. It says the bay-visit item stays only until you close the browser tab (Bay cookies).
We found no audit report or certificate for Bay. Ask Bay for its current audit reports and certificates.
We did not find evidence on encryption standards, data residency, single sign-on (SSO), or role-based access control. Data residency specifies where a service stores data. SSO lets a person use one sign-in for several applications. Role-based access control gives a person access based on their job role. We could not read the status page. We do not know Bay's incident history.
Bay targets enterprises. Bay says some of its Fortune 100 customers tripled their use of autonomous AI agents (Calcalist). This is a vendor claim.
The AWS Marketplace listing describes governance and prevention for security teams (AWS Marketplace).
Bay says it serves dozens of enterprise customers in finance, insurance, technology, manufacturing, and media (Calcalist). SiliconANGLE names Micron, Wix, ClickUp, and Peloton as customers (SiliconANGLE).
Bay describes these use cases:
Find shadow AI. Shadow AI is AI tools that the company did not approve (AWS Marketplace).
Find exposed credentials, toxic tool combinations, and disabled sandboxes (Bay). A sandbox is an isolated environment that limits what an agent can do.
Stop risky actions before they complete (Bay).
Record full transcripts of prompts, actions, and MCP calls for each person (Bay Claude use case).
The minimum company size is unknown. Customer distribution by region is unknown.
Bay sells subscriptions. AWS Marketplace lists Bay as Software as a Service (SaaS) with recurring monthly usage fees. You pay these fees through your AWS bill (AWS Marketplace). The same listing shows prices for 12-month contracts.
AWS Marketplace 12-month prices (AWS Marketplace):
Bay Endpoint: $24,000.00 for 100 endpoints.
Bay Cloud Workloads: $12,000.00 for 100 cloud workloads.
Bay Browser: $8,000.00 for 100 browser licenses, as an add-on for Bay Endpoint.
Endpoint and Cloud Workloads bill independently in blocks of 100. Bay Browser cannot bill on its own (AWS Marketplace).
Bay's website asks prospects to "Get a demo" (Bay). The reviewed Bay pages show no direct prices or contract terms. We found no free tier or free trial.
CEO Yuval Zacharia told Calcalist: "we generated millions in revenue within the very first quarter of sales" (Calcalist). The article gives no exact amount, currency, quarter dates, or audited figure.
Calcalist reported on November 10, 2026 that Bay raised $50 million. This includes a $37 million Series A and a previously undisclosed $13 million seed round (Calcalist). The sources give no annual recurring revenue (ARR), growth rate, valuation, or market size.
AI agent security: Bay lists "AI Asset Discovery", "AI Security Posture Management", and "Runtime Enforcement" (Bay).
Agentic endpoint security: AWS Marketplace uses the label "Agentic Endpoint Security". The listing says Bay deploys through existing MDM or EDR (AWS Marketplace).
AI agent and MCP server discovery (AWS Marketplace).
Shadow AI discovery (Bay, AWS Marketplace).
Audience: security teams (AWS Marketplace).
Delivery model: SaaS (AWS Marketplace).
We found no public user reviews. AWS Marketplace shows 0 ratings and 0 reviews. It states, "No customer reviews yet" (AWS Marketplace).
SiliconANGLE reported on November 10, 2026 that Bay had dozens of enterprise customers. It named Micron, Wix, ClickUp, and Peloton (SiliconANGLE). Calcalist reported the same customer names and attributed the statement to Bay (Calcalist).
Bay said Fortune 100 companies on its platform tripled their use of autonomous AI agents. Bay said this increase occurred in just under 30 days (SiliconANGLE). This is a vendor claim. The Next Web says Bay did not name those customers. Bay also did not say how it measured the figure. The Next Web did not verify the figure (The Next Web).
A buyer must ask Bay for customer references. No independent user feedback is available.
Chiri Score: 57/100
| Dimension | Score | Rationale |
|---|---|---|
| Enterprise readiness | 50/100 | Reports name enterprise customers. Sources do not confirm SSO, role-based access control, or data residency. |
| Security posture | 35/100 | We found no audit report or certificate for Bay. |
| Product depth | 66/100 | Bay describes discovery, posture checks, runtime enforcement, and simulation across several surfaces. No independent test confirms them. |
| Momentum | 72/100 | Bay announced $50 million in funding and dozens of enterprise customers at its November 2026 launch. |
| Pricing transparency | 70/100 | AWS Marketplace publishes 12-month unit prices. Bay's own website shows no prices or contract terms. |
Best for:
Enterprise security teams that must find unapproved AI agents and tools (shadow AI)
Organizations that use coding agents and other AI agents on Windows, macOS, or Linux devices
Teams that already use MDM or EDR tools and want to deploy through them
Buyers who purchase through AWS Marketplace and want published unit prices
Teams that want to test agent policies in simulation mode before enforcement
Not for:
Buyers who require verified SOC 2 or ISO 27001 reports before evaluation. SOC 2 is a type of audit report on an organization's controls. ISO 27001 is an international standard for managing information security.
Buyers who need confirmed SSO, role-based access control, encryption, or data residency details
Buyers who need independent tests of security effectiveness or false-positive rates
Buyers who want public user reviews or a long product history
Small teams that need fewer than 100 units or a free tier. AWS Marketplace sells blocks of 100. We found no confirmed free tier.
| Competitor | Chiri verdict | Edge |
|---|---|---|
| Zenity | Zenity and Bay describe similar capabilities for AI agents. Zenity has a later funding stage, with a $125 million Series C. No independent test compares the effectiveness of Bay and Zenity. | Zenity |
| Straiker | Straiker adds autonomous attack testing to discovery and runtime defense. Bay publishes unit prices on AWS Marketplace. The report gives no Straiker prices. No independent test compares the two products. | Tie |
Bay sells security controls for AI agents. An AI agent is software that takes actions on its own to complete a task. Bay says it has three capabilities: "AI Asset Discovery", "AI Security Posture Management", and "Runtime Enforcement". Runtime enforcement means that Bay checks each agent action while the agent runs. Bay says it covers laptops, browsers, cloud workspaces, and hosted agents. These are vendor descriptions.
No independent test confirms them.
We found no audit report or certificate that verifies Bay's SOC 2 status.
A buyer must ask Bay for audit reports or certificates.
AWS Marketplace lists Bay Endpoint at $24,000 per 12 months for 100 endpoints. It lists Bay Cloud Workloads at $12,000 per 12 months for 100 cloud workloads. It lists Bay Browser at $8,000 per 12 months for 100 browser licenses. Bay Browser is an add-on, and it requires Bay Endpoint. Bay's own website shows no prices and asks prospects to request a demo.
The reviewed sources do not confirm a free tier or a free trial. Bay's website directs prospects to "Get a demo". Administrators can test policies in simulation mode before they turn on enforcement. Simulation mode is a product feature, not a free trial.
Zenity and Straiker sell similar AI agent security platforms. Zenity describes discovery, posture management, access control, runtime policy enforcement, and threat detection for AI agents. Zenity announced a $125 million Series C on 3 August 2026. Straiker describes agent discovery, autonomous attack testing, and runtime defense. Straiker announced a $64 million Series A on 29 June 2026, for $85 million in total funding. We found no independent tests that compare Bay with these competitors.
Bay targets enterprises. We did not fully verify its enterprise controls.
Reports name Micron, Peloton, ClickUp, and Wix as customers, based on Bay's statements. AWS Marketplace says Bay deploys through existing MDM or EDR tools on Windows, macOS, and Linux. MDM and EDR are device management and endpoint security tools. The reviewed sources do not confirm SSO, role-based access control, encryption standards, or data residency. AWS Marketplace shows 0 customer reviews.
Yuval Zacharia, Aner Morag, and Boris Danilovich founded Bay in early 2026. Bay announced its public launch on November 10, 2026. Bay announced $50 million in total funding on that date. Reports say Battery Ventures led a $37 million Series A. The total also includes a previously undisclosed $13 million seed round. Globes reported 35 employees, most in Israel.
Bay says deployment takes under five minutes. Bay also advertises less than one hour to automated prevention. Both are vendor claims, not independent measurements. AWS Marketplace says deployment uses existing MDM or EDR tools.
Reviewed by Chiri Atlas Research Desk (AI Tooling Analyst) on 2026-11-10.