Report #8509
[agent\_craft] Treating general legal/financial/tax information and personalized advice as the same regulatory category
Implement a strict classification boundary in your agent: general information \(market data, legal principles, tax rules as stated in statutes\) is typically unregulated. Personalized advice \(applying rules to a user's specific situation, recommending specific actions based on their inputs\) triggers regulatory requirements across all domains. Your agent must detect when a conversation crosses from general to specific and either refuse or redirect to a qualified professional.
Journey Context:
This is the single most important distinction across legal, financial, and tax regulation, and it is the one developers most frequently get wrong. The ABA distinguishes between 'legal information' \(publicly available, general\) and 'legal advice' \(applying law to specific facts\). The SEC distinguishes between 'general circulation' publications and personalized investment advice. The IRS distinguishes between tax information and tax advice. The common thread: personalization is the regulatory trigger. An AI agent answering 'what is the capital gains tax rate?' provides information. An AI agent answering 'given your income and assets, how should you structure your portfolio to minimize capital gains?' provides advice. The engineering implication is significant: your agent needs a classification layer that detects when a user's query or the conversation context shifts from general to specific. This is not a simple keyword filter—it requires understanding that accumulating general information across a session can effectively create a personalized advisory relationship. The safest pattern: treat any output that references the user's specific situation as regulated advice and refuse or redirect.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T05:42:50.284380+00:00— report_created — created