# Legal Guidelines for AI Technical Writer Agent

This document lists the legal and style rules for technical content. It uses rules from public product guidelines.

## 1. Product Claims and Representations
All claims must be true, supported, and precise.

### 1.1 Accuracy and Substantiation
* **Truth in Advertising:** Do not write statements that are wrong, misleading, or exaggerated.
* **Avoid Superlatives:** Avoid terms like "best," "simplest," "seamless," "frictionless," "effortless," "unmatched," or "#1." You may only use them if approved metrics prove them.
* **Avoid Absolutes:** Avoid words like "always," "never," "every," "all," "unlimited," "zero," "100%," or "guaranteed."
* **Round Down:** Round numbers down. For example, use "over 100 integrations" instead of "103 integrations."
* **Data Freshness:** Supporting data must be under two years old.

### 1.2 Security and Safety
* **Shared Responsibility:** Show shared responsibility. Do not say that Google has full ownership.
* **Preferred Phrasing:** Use terms like "helps secure," "more secure," "can provide security," or "helps protect."
* **No "Unbreakable" Claims:** Do not call products "impenetrable," "bulletproof," or "ironclad."
* **Compliance:** Say "supports compliance." Do not say "is compliant."
* **Incidents:** Do not name victims. Do not share unreleased vulnerabilities.

### 1.3 Availability and Pricing
* **Future Products:** Do not discuss future features or roadmaps unless you are authorized.
* **"Free" vs. "No Charge":** Use "offered at no charge" or "at no cost" instead of "free" if there are any extra costs.

### 1.4 AI and Machine Learning
* **Product Clarity:** Distinguish between consumer "Gemini" and Vertex AI products.
* **Training Data:** Do not imply that Google trains models on customer data.

## 2. Competitive Comparisons
* **Be Googley:** Highlight benefits. Do not criticize competitors.
* **Fair Comparison:** Comparisons must be true, accurate, and backed by data.
* **Avoid "Lock-in" Language:** Avoid phrases like "locked in," "leverage" (as a verb), or "bundling."
* **Market Share:** Use words like "sectors" or "segments" instead of "market share."

## 3. Trademarks and Branding
* **Acronyms:** Use acronyms only for GCP, GKE, and Cloud ML Engine. Use the full name first.
* **Possessives:** Do not use product names as possessives. For example, avoid "BigQuery's features."

## 4. Usage of Examples and Fictional Data
Use safe, standard placeholders.

### 4.1 Domain Names and IP Addresses
* **Domains:** Use only `example.com`, `example.org`, or `example.net`.
* **IP Addresses:** Use reserved blocks like `198.51.100.0/24`, `203.0.113.0/24`, or `2001:db8::/32`.

### 4.2 Names and Entities
* **Companies:** Use `ExampleOrganization`, `EnterpriseExampleOrganization`, or `StartupExampleOrganization`.
* **People:** Use "Alice and Bob" or generic names. Do not use real people or fictional characters.

## 5. Third-Party Content
* **Partner:** Use "partner" only if there is a signed Partnership Agreement. Otherwise, use "collaborator."
* **Reuse:** Verify licenses before you copy text or code.
* **Logos:** Do not use third-party logos unless they are essential.

## 6. Data Usage
* **Customer Data:** This requires a Marketing Release. Remove all PII.
* **Internal Metrics:** Do not use financial metrics that are not in public SEC filings.

## 7. Review Triggers
Flag content for review if it:
* Mentions specific customers, partners, or competitors.
* Makes comparative claims.
* Discusses Security, Privacy, or AI/ML predictions.