The Shield360 AI Control Plane
What AI cost, who ran it, what data it touched and what was stopped — one record, in your environment or ours.
Two ways to use Shield360
One platform, reached two ways. Govern how your people use AI, or monitor the AI you ship to customers.
Your AI system of record
Shadow AI, personal accounts, coding agents, agents calling tools, spend no invoice explains: one record of all of it, with no proxy and no root certificate.
Unedited captures of the public demo. Synthetic data.
Four modules, one record
Four modules on one dataset, each available alone. Start with your biggest AI challenge.
Shadow AI by person and computer, with no root certificate and no proxy changes. Prompts carrying secrets or personal data are warned on, masked or blocked.
No root certificate Discovery needs no certificate and no proxy changes.
One policy The browser extension, the AI proxy and coding-agent hooks read the same organization-wide policy.
Explore Security resources:
Solution
For the CISO
Exposure, controls and agent access.
Read the page
Docs
Enforcement policy
Off, warn, mask or block, per kind of data.
Read the docs
Got a different challenge? Find your weakest area with the assessment.
Book a briefingSpend by person, machine, tool and model, from the invoice where a bill is connected and marked as an estimate where not.
Untracked spend, shown Spend no plan explains gets its own line, never hidden in another.
Paid for, unused Seats nobody holds, or nobody used in 30 days, are listed as exactly that.
Explore Finance resources:
Walkthrough
Costs, screen by screen
Every dollar, and the basis for it.
See the screens
Brief · PDF
The enterprise brief
One document to forward.
Download the brief
Got a different challenge? Find your weakest area with the assessment.
Book a briefingControl status for ISO/IEC 42001, the EU AI Act and NIST AI RMF from observed evidence, kept daily. Evidence, not a certification.
38 of 38 ISO/IEC 42001 Annex A controls, each with a Statement of Applicability entry.
Never edited The audit log records reads as well as changes; the database refuses to update or delete it.
Explore Compliance resources:
Guide
Governing agentic AI
Six pillars, and where each stops.
Read the guide
Checklist
AI governance assessment
Eighteen questions, scored in your browser.
Start the checklist
Got a different challenge? Find your weakest area with the assessment.
Book a briefingCoding-agent analytics for Claude Code, Cursor and Codex, data guards, and approved models enforced at the AI proxy.
By tool and repository Lines written by AI are reported by tool and repository, never as a measure of a person.
Approved models only The AI proxy can warn on, or refuse, a model outside the approved list.
Explore Developer resources:
Docs
Coding-agent analytics
What is recorded from each coding agent.
Read the docs
Architecture
What crosses the boundary
What listens, and what leaves a machine.
Read the architecture
Got a different challenge? Find your weakest area with the assessment.
Book a briefingEnforcement points
Off, warn, mask or block per kind of sensitive data, from one organization-wide policy.
Checks prompts on eleven AI chat sites; allows, warns or blocks the rest.
How it worksBy base URL, no root certificate. Approved providers and models only.
How it worksClaude Code, Cursor and Codex: prompts, secret files, injection.
How it worksA token per agent; deny-by-default grants, read on every call.
How it worksYour app asks before it calls a model. The text is never stored.
How it worksNo TLS interception: internal apps are covered through the AI proxy or the guard API. Other AI use is still discovered and reported.
The platform
Each a working area of the product today, not a roadmap item.
Deployment
The same software either way: you choose where the data lives.
Self-hosted · air-gapped
Hosted SaaSBuilt to be checked
Built into the product, not left to a setting.
No root certificate And no proxy changes, on any machine.
7 years Every block and warning, kept by the database.
Tamper-resistant Audit log entries can’t be edited or deleted.
Safe by default If the policy can’t be reached, nothing extra is collected.
Either. Install it in your own environment — on Docker, or on Kubernetes with manifests we provide, against your own ClickHouse and Postgres, including fully air-gapped — or use the hosted service, where each organization gets its own database. Both are the same software with different configuration, so a hosted pilot can move in-house for production.
Not unless you choose to put it there. The collector installs no certificate, changes no proxy setting and opens no listening socket. It reads files the AI tools already write and tables the operating system already keeps, then sends what it found outbound over TLS. The one part that does sit in the path is the optional AI proxy, and only for applications you point at it by base URL. It needs no root certificate, and traffic you do not route through it never reaches it.
Data protection runs at the point of use. The browser extension, in managed Chrome and Edge, allows, warns about or blocks the AI sites on its list, and on eleven AI chat sites, ChatGPT, Claude, Gemini and Microsoft Copilot among them, can warn about, mask or block a prompt carrying a secret or personal data before it is sent. Hooks inside coding agents can warn or block. Where you want enforcement on the wire, the optional AI proxy applies the same policy and can refuse models outside your approved list, and your own applications can ask the guard check API before they send text to a model. One organization-wide policy drives all of them.
Data protection reads a prompt before it is sent, so it can act on it. Storing it is a separate choice, and it is off by default: no prompt, response or tool content is kept unless an administrator turns that on. What is always kept is the model, token counts, cost, the guard's decision and the kinds of sensitive data found. An environment can only store less than its organization, and secrets are always scrubbed from anything stored.
The hosted service has published plans: Free, Team and Business, each with apps, computers, spans and retention included, and add-ons for more. Paid plans start with a 14-day trial, no card needed. See pricing. Self-hosted and air-gapped deployments are Enterprise: book a briefing and we’ll price it with you. Or start free on the hosted service.
Ready for a conversation?
A working session on the live product. We say plainly what Shield360 can and cannot answer.
Booked through cal.com. Pick a starting point, or just a time.
Book a briefingOr start from a question
