Last updated: November 3, 2026. Testing track: Documentation Reviewed.
Island is an enterprise browser vendor that has expanded into an AI control plane covering browser, extension, desktop, network and agent activity. Its Enterprise AI offerings, launched March 17, 2026, add AI usage governance with redaction, a multi-provider AI chat inside the browser, governed agents and an MCP gateway. The strongest controls assume users work in the Island browser, which is a larger change than an extension or API product.
Score: 6.7 / 10.
Initial review November 3, 2026.
This review is based on Island’s public documentation, product pages, pricing pages and press coverage as of November 2026. We have not had lab access or a vendor demo.
Scores use the same published rubric and formula as every review on this site. Sources are listed at the end of the review.
Coverage breadth
Detection accuracy
Deployment friction
Policy & control depth
Framework alignment
Support & documentation
20%
20%
15%
15%
10%
10%
10%
8
7
5
9
4
5
7
Browser, extension, desktop, network, coding agents and MCP are covered, but connected-app governance is absent.
DLP, redaction and a prompt-injection detector are documented without public benchmarks.
Full value requires rolling out a new browser; the extension option reduces but does not remove that friction.
Very granular browser, data, tenant, agent and per-tool controls including redaction and response interception.
Mentions EU AI Act support and SOC 2 or ISO 27001 exports but no AI framework mappings.
Quote only from Island; figures exist only in a reseller’s public-sector price list.
SIEM integrations documented through partners and solid vendor learning content, but public product docs and support tiers are not available.
Score
8
Notes
Browser, extension, desktop, network, coding agents and MCP are covered, but connected-app governance is absent.
Score
7
Notes
DLP, redaction and a prompt-injection detector are documented without public benchmarks.
Score
5
Notes
Full value requires rolling out a new browser; the extension option reduces but does not remove that friction.
Score
9
Notes
Very granular browser, data, tenant, agent and per-tool controls including redaction and response interception.
Score
4
Notes
Mentions EU AI Act support and SOC 2 or ISO 27001 exports but no AI framework mappings.
Score
5
Notes
Quote only from Island; figures exist only in a reseller’s public-sector price list.
Score
7
Notes
SIEM integrations documented through partners and solid vendor learning content, but public product docs and support tiers are not available.
Controls apply at the browser, where Island can redact prompts and intercept AI responses before they render, plus separate corporate and personal AI tenants (Island AI page; Island press release, March 2026).
Provides a sanctioned multi-provider AI chat inside the browser, so organizations can offer approved models rather than only block (Island press release, March 2026).
Extends beyond the browser to desktop, network, coding agents and an MCP Gateway with per-tool permissions (Island AI page).
Very well funded, with a $400 million Series F at a $6.4 billion valuation in September 2026 (Reworked). Logs reach common SIEMs through a generic webhook integration (Google SecOps docs).
The deepest controls require adopting the Island browser, a bigger change management project than an extension or API product (Island press release; AWS Marketplace listing).
Pricing is not published by Island, and the only public figures come from a reseller's government price list showing a $250,000 annual console SKU (G-Cloud 15 price list).
No published mappings to NIST AI RMF, OWASP, ISO 42001 or MITRE ATLAS (not documented). Detection quality claims, including prompt injection detection, have no published benchmarks (Island AI page). Microsoft 365 and Google Workspace connected-app governance is not documented.
Enterprises already using or planning an enterprise browser for contractor access, VDI replacement or SaaS data protection, that want AI usage governance, redaction, a sanctioned AI chat and agent controls in the same platform. Larger organizations with budget for a platform purchase get the most value.
Smaller organizations or teams that want to keep users in Chrome or Edge with a lightweight extension, need transparent pricing, or need published framework mappings and AI vendor risk assessments. Alternatives include Harmonic, Nudge or Prompt Security for extension-based AI usage control, Microsoft Purview for Microsoft 365 centric estates, and Lakera for application-layer prompt defense.
Not published on island.io (request a quote); the AWS Marketplace listing shows custom pricing. A UK G-Cloud 15 price list from reseller Somerford Associates shows Island Managed Browser at $200 per user per 12 months and Island Management Console at $250,000 per customer per 12 months; AI-specific SKUs are not listed.
See how it compares in our Best AI DLP Tools 2026 ranking. The poor-fit section above names alternatives for buyers it does not suit, and our methodology explains how every product is scored.
If Island grants lab access, we would run the following scenarios and upgrade this review to Lab Tested.
The standard sensitive-data set (PII, PHI, payment data, secrets, source code) sent through the product's protected AI tools.
Inventory accuracy against a known set of sanctioned and unsanctioned AI tools.
Block, warn, redact and allow behaviors against a configured policy.
What is logged, what is not, and retention behavior.
Microsoft Entra ID and Okta where supported.
Time from access to enforced policy in a representative environment.
Added latency on standard prompt sizes.
Based on public documentation. “Not documented” means we could not find it in public material, not that the product lacks it.
| Capability | Island |
|---|---|
| Inline prompt DLP | Yes. Inspects prompts, files and tool calls inline and redacts sensitive data before it reaches any AI provider. |
| Shadow AI discovery | Yes. Discovers AI tools, agents, MCP servers, Bash commands and models, with risk scoring. |
| Blocking unapproved AI tools | Yes. Governs unsanctioned AI apps and can redirect users to sanctioned alternatives. |
| File upload inspection | Yes. Files are inspected inline with DLP; browser upload controls apply. |
| Secure AI workspace (sanctioned chat) | Yes. AI Browser adds AI chat to the Island browser with multiple connected providers (for example ChatGPT, Copilot, Claude) and enterprise data via RAG. |
| Built-in AI vendor risk scoring | Partial. Risk scoring for discovered AI tools and for browser extensions; a vendor-level AI risk assessment methodology is not documented. |
| Microsoft 365 connected AI app review and control | Not documented. No Microsoft 365 OAuth app review or revoke documented. |
| Google Workspace connected AI app review and control | Not documented. No Google Workspace connected-app governance documented. |
| Governance scorecard, attestation and board reporting | Partial. Unified dashboard with policy audit trails and content-level audit records exportable for SOC 2 and ISO 27001; no board-ready AI governance scorecard documented. |
| AI agent and MCP security | Yes. MCP Gateway with per-source policy, identity, audit and threat intelligence; Agentic Identity issues just-in-time credentials for agents. |
| Prompt injection defense | Yes. Prompt-injection detector and LLM-as-a-judge evaluator inspect prompts and tool calls inline. |
| AI red teaming | Not documented. No AI red teaming offering described. |
| SIEM integration | Yes. Generic SIEM webhook integration in the Management Console, documented by Google SecOps, with regional US and EU endpoints. |
Chromium-based enterprise browser deployed as SaaS-managed software, or the Island Extension layered on existing browsers; Island Desktop endpoint component; Island Network for inline inspection; MCP Gateway for agent tool access; ZTNA connectors for internal apps. Typical rollout time is not published.
Web AI apps in the Island Enterprise Browser and, via the Island Extension, other browsers; desktop and thick-client AI apps via Island Desktop; inline network inspection via Island Network; coding agents (Cursor, Windsurf, Cline), desktop and autonomous agents, homegrown apps on LangChain or LangGraph; MCP servers, skills and browser extensions on endpoints.
Not entirely. The Island Extension applies controls in other browsers and Island Desktop covers desktop AI apps, but the full feature set is built around the Island Enterprise Browser.
Yes. Island says it redacts sensitive data before it reaches any AI provider and can intercept AI responses before they render.
Yes. Its MCP Gateway applies per-source policy, identity and audit, and agents run with admin-set permissions and human approval controls.
Island does not publish pricing. A UK G-Cloud reseller list shows $200 per user per year for the managed browser and $250,000 per year for the management console.