of organizations that reported an AI-related breach lacked proper AI access controls.
IBM Cost of a Data Breach 2025
New
PexLens Enterprise AI Browser: the enterprise AI browser with security built in.
See what’s inside →
Where AI enters your business
Three entry points. Three PexLens controls. One policy.
Sensitive data and risky code reach AI through more than a typed question. Select an entry point to see how PexLens covers it.
01
PexLens TrustGateAI DLP · AI data loss prevention
AI prompts, files and conversations
Control what employees can share with AI, with which AI tools, and under what policy, across prompts, files, uploads and AI-connected workflows.
Employees do not only send text prompts. They paste confidential documents, upload spreadsheets, share source code, submit customer records, attach internal policies and ask AI to analyze sensitive business data.
The risk is not limited to one AI platform. It applies to public AI chat interfaces, enterprise AI assistants, AI features embedded in SaaS applications, browser-based AI tools and locally installed AI applications.
PexLens helps organizations apply consistent controls across these interactions:
Detect sensitive data before it is submitted to an AI service.
Apply policies based on data type, user role, destination and business context.
Prevent source code, credentials, customer data, financial records and internal documents from reaching unapproved AI tools.
Give security teams visibility into AI usage without disrupting legitimate productivity.
Support safe use of approved AI tools while reducing shadow AI.
WHY IT MATTERS
Sensitive information can leave the organization through prompts, uploads, copied text and AI-connected workflows, not only through a typed question.
PEXLENS VALUE
PexLens helps organizations use AI confidently by controlling what employees can share, with which AI tools, and under what policy.
Explore PexLens TrustGate →$5 / device / month
02
PexLens Enterprise AI BrowserSEB · Secure enterprise browser · Chromium-based
The browser where AI work happens
The enterprise AI browser with security built in. It turns the browser from an unmanaged AI risk surface into a controlled, observable workspace.
The enterprise AI browser with security built in. Much of modern AI usage happens in the browser: AI chat tools, AI-enhanced SaaS applications, browser extensions, web-based agents and AI features built into business applications.
The browser is also where traditional security controls often lose visibility. Employees may paste sensitive information into an AI interface, upload files, use personal AI accounts, install unapproved extensions or connect AI agents to business systems.
PexLens Enterprise AI Browser helps secure this layer with:
Visibility into AI tools, extensions and AI-powered SaaS usage.
Control over which AI tools employees can access.
Real-time inspection of prompts, uploads, downloads and copy/paste activity.
Data loss prevention (DLP) for AI interactions.
Extension risk scoring, approval workflows and permission monitoring.
Protection against malicious or overly privileged browser extensions.
WHY EXTENSIONS MATTER
Browser extensions can request broad access to page content, form data, cookies, session tokens and enterprise applications. A compromised extension can become a silent channel for data theft or account takeover.
PEXLENS VALUE
PexLens turns the browser from an unmanaged AI risk surface into a controlled, observable workspace.
Extensions · risk scoringILLUSTRATIVE
AI page summarizer · reads all sites and cookies✕ Blocked
AI writing helper · form data access⚠ Needs approval
Corporate password manager · approved✓ Allowed
03
PexLens SupplytrustSSCS · Software supply chain security · SCA · AI model scanning · AI-BOM
Packages, extensions and AI tools
Verify packages, AI models, extensions, plugins, MCP servers and AI tooling before they enter your environment, whether a developer, an AI agent or a non-technical employee installs them.
The third entry point is the software supply chain. Developers have always installed packages, libraries, IDE extensions and build tools. AI changes the scale and speed of that risk: developers may accept an agent’s suggested dependency, and non-technical employees may use AI agents to build internal tools, automate workflows or install packages without understanding the security implications.
AI agents can also download and install dependencies automatically. In some cases the human checkpoint disappears entirely: an agent selects a package, runs an installation command and executes install scripts before anyone reviews the change.
PexLens Supplytrust helps protect this surface by:
Scanning packages, extensions, plugins, containers and AI tooling before they are broadly adopted.
Detecting malicious, suspicious or typosquatted components.
Identifying unexpected permissions, network behavior, persistence mechanisms and credential access.
Comparing new versions against known-good releases to detect supply-chain drift.
Evaluating MCP servers, agent skills, plugins and AI-powered extensions.
AI model scanning: checking models pulled from Hugging Face and other model hubs for unsafe serialization such as pickle, embedded code execution, unexpected network calls and tampered weights, before a model is loaded.
Homebrew and system packages: covering Homebrew formulae and casks alongside npm, PyPI, Maven and RubyGems, so tools installed on macOS and Linux developer machines get the same checks.
Organization-based policies: approved sources, model hubs, registries and taps per team, device group or project, with allow, quarantine, block or require-approval decisions.
Giving security teams policy controls to allow, quarantine, block or require approval.
AI models
Hugging Face and model hubs
A model file can run code the moment it loads. Supplytrust scans models before they reach notebooks, pipelines or production.
Homebrew · org policy
Formulae, casks and taps
Allow only approved taps and casks per team, and stop unvetted tools landing on developer laptops.
WHY IT MATTERS
Employees outside engineering are now building tools with AI. Research on AI-agent usage found non-engineering users running code, builds, package installs and cloud or cluster commands; in one organization, 12% of non-engineering users installed packages through agents.
Source: Sysdig AI usage researchPEXLENS VALUE
PexLens protects the components that enter your environment, whether installed by a developer, an AI agent or a non-technical employee building an internal tool.
Explore PexLens Supplytrust →From $0 / month
AI agent · built by a finance analyst · terminal
agent › I’ll add an HTTP client for the report sync.
$
✕ BLOCKED by PexLens Supplytrust
typosquat of “requests” · install script reads ~/.aws and calls an unknown host
policy: block malicious packages · decision logged
typosquat of “requests” · install script reads ~/.aws and calls an unknown host
policy: block malicious packages · decision logged
agent › Switching to the verified package.
$ pip install requests ✓ ALLOW
agent › Adding MCP server “sheets-sync”…
⚠ REQUIRES APPROVAL new MCP server · broad file access
agent › Loading model for the forecast notebook…
>>> from_pretrained("acme-labs/fin-forecast-7b")
✕ MODEL BLOCKED pickle payload calls os.system · Hugging Face source
$ brew install --cask ai-notes-sync
⚠ ORG POLICY cask not on Finance approved list · request sent to IT
12%of non-engineering users installed packages through AI agents in one organizationSysdig AI usage research
42%of non-engineers using agents had them run code, builds, package installs or cloud commandsSysdig AI usage research
Why AI security, why now
AI adoption is outpacing AI security.
Employees paste data into AI tools, developers let AI pick dependencies, and attackers target both. Most organizations still have no control point in between.
added to average breach costs for organizations with high levels of shadow AI. One in five reported a shadow AI breach.
IBM Cost of a Data Breach 2025
of enterprise employees using generative AI admit entering confidential information into public AI tools.
TELUS Digital survey, 2025
new open source malware packages found in one quarter across npm, PyPI and Hugging Face, up 140%.
Sonatype Open Source Malware Index, Q3 2025
The PexLens platform
Three products. One policy engine.
PexLens TrustGate, PexLens Enterprise AI Browser and PexLens Supplytrust share one policy engine and one view of risk, so a rule you write once applies to the prompt an employee types, the browser they work in and the package a developer installs.
PexLens TrustGateUSE AI
Protect every prompt
AI DLP · AI data loss prevention for prompts and responses
Inspects prompts and responses on the device across ChatGPT, Claude, Gemini, AI search and local developer tools.
Detects source code, secrets, API keys and PII across 30+ categoriesAllow, redact, block or log by policyKeeps teams on AI instead of banning it
$5 / device / monthSee TrustGate →
Secure AI browsingWORK WITH AI
PexLens Enterprise AI Browser
SEB · Secure enterprise browser for AI
The enterprise AI browser with security built in. Chromium-based, with control over AI-assisted work, sessions and sensitive data.
Allow-list the AI models and instances you approveClipboard and data-leakage shieldingAI usage dashboards and audited prompt history
From $12 / seat / monthSee the PexLens Browser →
PexLens SupplytrustBUILD WITH AI
Secure what AI builds with
SSCS · Software supply chain security with SCA and malware blocking
Real-time protection for npm, PyPI, Maven, RubyGems and Homebrew, AI models from Hugging Face, browser and IDE extensions, AI coding assistants and agent components.
Blocks malicious and typosquatted packages at installScans AI models before they load, with org-based source policiesPriced by unique package versions, never per developer
From $0 / monthExplore →
Where AI risk enters
Six ways AI creates risk. One platform that covers them.
01 · SHADOW AI AND DATA LEAKS
Sensitive data pasted into AI tools
Code, credentials and customer data leave through prompts and uploads, often from personal accounts.
TrustGate · PexLens Browser 02 · PROMPT INJECTIONHidden instructions in content AI reads
Web pages, files and tool responses can steer assistants into leaking data or taking unintended actions.
TrustGate · Supplytrust 03 · UNAPPROVED MODELSAI services nobody approved
Teams adopt new chatbots, copilots and AI search tools faster than policy can keep up.
PexLens Browser · TrustGate 04 · MALICIOUS PACKAGES AND MODELSDependencies and models AI pulls in
Assistants recommend packages that are malicious, typosquatted or do not exist yet, and models downloaded from public hubs can run hidden code when loaded.
Supplytrust 05 · RISKY AI EXTENSIONSCopilots and extensions with deep access
Browser and IDE extensions read pages, code and credentials, and can turn hostile through a single update.
Supplytrust · PexLens Browser 06 · AGENTIC AI AND MCPAgents that act on their own
Coding agents and MCP servers extend what AI can reach and do, with your credentials.
Supplytrust · TrustGateHow PexLens works
From AI signal to enforced decision, in real time.
See every AI touchpoint
AI tools, models, extensions, packages and agent components across devices, browsers and CI.
Detect AI signals early
Sensitive data in prompts, risky AI components and malicious packages, before they reach production.
Evaluate risk in context
Signals meet dependency, identity and policy context, and return allow, redact, alert or block.
Audit every decision
A complete record of AI use and enforcement for security reviews, auditors and incident response.
Built for the teams that own AI risk
One platform. Four teams. One set of rules.
Security leadership
Approve AI with confidence: one view of AI risk and one policy across prompts, browsers and code.
AppSec / DevSecOps
Stop malicious packages and risky AI tools in IDEs, CI and agents without slowing developers.
IT & Endpoint Security
Roll out the PexLens Browser and TrustGate with zero-touch deployment and central browser policies.
GRC & Privacy
Turn your AI policy into enforced controls, with audit trails for every AI decision.
Simple, predictable pricing
Start free. Scale with AI adoption.
Works with any plan, or on its own.
Core, Govern and Command plans. Minimum 50 seats, save 20% on annual contracts.
Free for npm and PyPI. Priced by unique package versions, never per developer.
What is AI security?
AI security protects how an organization builds and uses AI: the data employees share with AI tools, the AI-related components and extensions developers install, and the agents that act on their behalf.
What is PexLens Enterprise AI Browser?
PexLens Enterprise AI Browser is the enterprise AI browser with security built in, for secure, managed AI-powered browsing. The PexLens Browser is Chromium-based and uses the same policy engine as PexLens TrustGate. Plans start at $12 per seat per month (Core), with Govern at $15 and Command at $21.
How does PexLens stop sensitive data leaking into AI tools?
PexLens TrustGate inspects prompts on the employee device and detects source code, secrets, API keys and personal information before they reach an AI service. Depending on your policy it allows, redacts or blocks the content and logs the decision.
Do I need PexLens TrustGate if I use the PexLens Browser?
They work together. The PexLens Browser controls AI sessions, models, clipboard and browser policy, and TrustGate inspects prompts and responses on the device, including in browsers and local tools you already use.
How does PexLens Supplytrust relate to AI security?
AI assistants and agents install packages, extensions and tools on your behalf. PexLens Supplytrust, our software supply chain security (SSCS) product, checks each one at install time, so AI-driven development cannot bring malicious code into your environment.
Does PexLens scan AI models and Homebrew packages?
Yes. PexLens Supplytrust scans AI models from Hugging Face and other model hubs for unsafe serialization, embedded code and tampering before they are loaded, and covers Homebrew formulae and casks alongside npm, PyPI, Maven and RubyGems. Organization-based policies let you approve sources, model hubs and taps per team, device group or project.
Can we define our own AI security policies?
Yes. AI guardrails are designed around organization-defined rules, so your team decides which AI-related signals are approved, need review or should be blocked.
Can PexLens enforce AI usage policies?
Yes. The same policy engine can allow, warn on or block AI tools and AI-related components, and PexLens TrustGate applies policy to prompts and responses.
Where does AI enter a business?
In three places: the prompts, files and conversations people share with AI tools, the browser where AI work happens, and the packages, extensions, models and AI tools teams install. PexLens TrustGate, the PexLens Enterprise AI Browser and PexLens Supplytrust secure each one under one policy.
Who is PexLens for?
PexLens is for security leaders, engineering organisations, IT teams and compliance and privacy teams in organisations that want to adopt AI safely and protect the software they build and use.
How do we get started with PexLens?
Request a demo. The PexLens team will review the AI tools, browsers and package ecosystems you use and plan a pilot with your security and engineering teams.