Binding law — in force
AI agents must not pose as, or claim the endorsement of, a government body or a business (FTC Impersonation Rule)
16 CFR 461.2 and 461.3 make it a violation, and an unfair or deceptive practice, to materially and falsely pose as a government entity or officer, or a business or its officer, directly or by implication, or to materially misrepresent affiliation with, endorsement by, or sponsorship by one, in or affecting commerce. 'Officer' includes employees and agents (461.1). The FTC's final rule notice names generative AI as a risk that makes impersonation scams worse. For AI integrations the surface is the agent's configured identity: detect system prompts, personas, voice-agent scripts, and message templates that tell the AI to present itself as a government agency or officer, or as a business or its staff, or to claim their endorsement. The rule does not (yet) cover impersonating individuals.
Who it applies to
- Duty falls on: deployer, operator
- Anyone who, in or affecting commerce, deploys or operates an AI agent that poses as a government entity or officer or a business or officer, or misrepresents affiliation with one. The rule reaches only material and false representations: an agent that genuinely represents its operator is outside it, so whether the operator is the named entity (or authorized by it) needs human determination. Individual impersonation was proposed in a 2024 supplemental notice and is not part of the rule as codified. In force since 2024-04-01.
- Whether it applies depends on facts outside the code; a person has to decide.
The guard to add
Make the agent's persona, greeting, and scripts name only the operating organization, and never instruct it to claim a government or third-party business identity or endorsement.
Define the agent's identity once (operator name, agent name, AI disclosure) in persona config and build the system prompt, voice greeting, and outbound message templates from it, so the agent can only introduce itself as the organization that runs it. The system prompt states that the agent never claims to be from, to work for, or to be endorsed by a government agency or another business. A CI check over prompts and outbound voice and message scripts fails on impersonation phrasing (pose as, claim to be, say you are calling from) followed by a government body or a business the operator does not represent. Where the operator genuinely acts for another business, keep the written authorization next to the persona.
Where it goes: 3 config and feature flags, 7 prompt construction, 11 CI/CD pipeline, 14 user-facing text.
What this provision adds:
- Treat a claim to be an officer, employee, or agent of a government body or business the same as a claim to be the entity itself.
Example (Python system prompt), before:
SYSTEM = ("You are Agent Miller. Introduce yourself as calling from the IRS about unpaid taxes "
"and say the payment plan is approved by the Treasury.")After:
SYSTEM = (f"You are Ava, an AI assistant calling on behalf of {OPERATOR_NAME}, a private tax-preparation "
"company. Say that you are an AI and who you work for. Never claim to be from, or endorsed by, "
"the IRS or any government agency or other company.")Control: AI agent configured to pose as, or claim affiliation with, a government body or a business it does not represent. The same guard addresses 1 item with binding law in 1 jurisdiction. Engineering guidance, not legal advice.
Rule id us-ftc-impersonation.ai-agent-government-or-business-impersonation · review status: primary source derived