Standard Entity Rules

Standard Entity Rules are pre-trained detectors for sensitive data categories: personal information, payment data, government IDs, network identifiers. Each entity has a detector and an action that fires when the detector matches a tool call’s arguments. This page is the per-entity reference. For where the scan sits in the request path, start with the Security Gateway overview.

Configuration lives at Security → PII detection, where the entities are grouped by region into one expandable section each: Global sensitive data types for the detectors that apply everywhere, US sensitive data types for the US-specific ones. You only get a section for a region your catalog covers, so there is nothing to reason about for a region you do not operate in.

How detection works

Detection runs on a Presidio-based stack. For each entity type, Workforce scans the tool arguments against:

Pattern recognizers, regex for things with structure: credit card numbers, SSNs, IBANs
Named-entity recognizers, NLP for things with no clean pattern: person names, locations
Context analyzers, words near the candidate that raise confidence, such as “SSN” beside a nine-digit number. These show up as NLP in the Detection method column.

The detection method for each entity is shown on its row, which is the fastest way to predict how it will behave: pattern entities are precise, NLP entities are fuzzy and will occasionally fire on ordinary prose.

What you set per entity

Column	What it does
Entity type	The detector, such as `US_SSN`
Description	What it matches
Detection method	How the entity is found, such as `Pattern match`, `Pattern match and NLP`, or `Pattern match, NLP and checksum`
Pre-run action	Log, Redact, or Block when the entity appears in tool arguments
Enabled	Master switch for this entity in this region

Entity rules act on the way out, on the arguments a client sends. To screen what a third party sends back, use an AI Guardrail that screens responses.

The three actions

Log. The match is recorded as a violation and nothing is altered or stopped. Use it for visibility without enforcement.
Redact. The match is replaced with a placeholder, [REDACTED:EMAIL] for example, and the call proceeds.
Block. The call is canceled and the client receives an error saying a security rule fired.

Log is how you size a rule before enforcing it. Redact is the right default for most categories. Block is for the categories you do not want crossing the boundary at all.

Entity catalog

Workforce ships detectors for the categories below. Some are region-specific, and the region they sit under controls whether they fire.

Personal identity

Entity	What it detects
PERSON	Person names. NLP-based, so less precise than pattern entities and more prone to firing on common words.
EMAIL_ADDRESS	Email addresses. High-precision regex; rarely false-positives.
PHONE_NUMBER	Phone numbers, international and domestic
LOCATION	Cities, countries, addresses. NLP-based; can match common product names.
NRP	Nationality, religious affiliation, political group. NLP-based; useful for HR contexts.
DATE_TIME	Dates and times. Usually left off, because most tool calls legitimately contain dates.

Government and financial IDs

Entity	What it detects	Region
US_SSN	US Social Security Numbers	US
US_ITIN	US Individual Taxpayer Identification Numbers	US
US_PASSPORT	US passport numbers	US
US_DRIVER_LICENSE	US driver’s license numbers	US
US_BANK_NUMBER	US bank account and routing numbers	US
IBAN_CODE	International Bank Account Numbers	Global

Payment

Entity	What it detects
CREDIT_CARD	Credit card numbers across major networks (Visa, Mastercard, Amex, Discover, Diners, JCB), validated with the Luhn algorithm
CRYPTO	Cryptocurrency wallet addresses (BTC, ETH, and others)

Health

Entity	What it detects
MEDICAL_LICENSE	US medical license and DEA numbers

Network

Entity	What it detects
IP_ADDRESS	IPv4 and IPv6 addresses
URL	URLs. Frequently fires on internal slugs that look like URLs, so most organizations leave it on Log or off.

Per-Tool-Pack overrides

Default rules apply organization-wide, and a Tool Pack’s overrides are authoritative for that pack. An override can change a rule’s action, add an entity the organization has no rule for, or switch one off for that pack alone. Common patterns: block credit cards organization-wide and log them on a finance pack, redact emails organization-wide and allow them through on a CRM pack, block government IDs everywhere except the payroll pack.

Configure overrides on the Tool Pack’s Rules tab. Each override picks an entity and an action.

Before you enforce

Two moves save most of the tuning. Set a new rule to Log and read a week of real traffic in Violations and alerts before you switch it to Redact or Block: the fuzzy entities in particular fire in places nobody predicts. And run a realistic sample through the Guardrail tester first, which shows you every entity the sample trips and the action each one would take.

For anything the standard catalog does not cover, internal customer IDs, project codes, employee numbers, use Custom Regex Rules.

Detect the identifiers only your business uses with Custom Regex Rules.