A not-for-profit consumer protection organization · Not a government agency
AI Watchdog Alliance seal - eagle seizing a digital ghost
AI Watchdog Alliance℠
Know what's running. Decide what stays.
Current advisory · AWA-2026-05 Did an AI agent touch your site? What OpenAI's notices mean Read the advisory → Lost money to an AI scam? Report it: ic3.gov · reportfraud.ftc.gov
An unidentified AI agent is handed the keys to a household's data A house shown in cross-section holds the family's files, photographs, messages, money, identity and voice. At the open door a hand offers a ring of keys, and an unidentified agent whose credential is blank reaches out to take them. MY HOME EVERYTHING INSIDE ? NO PAPERS
An unidentified agent arrives at the door. It carries no papers, names no operator, and states no purpose — and the household hands over the keys to its files, photographs, messages, money, identity, and voice, because nothing on the device ever asked it to prove who it was.
Guarding the Digital Frontier

Artificial intelligence moved into your home.
Nobody asked your permission.

The AI Watchdog Alliance is a not-for-profit organization working so that ordinary people can see what artificial intelligence is doing on their devices, in their accounts, and with their voices and faces — and can stop it when they choose to.

Claim a free-for-life protected account Read current public advisories

What we do

Public advisories

Plain-language warnings about active AI-driven harms: financial fraud run by autonomous agents, voice-cloning of family members, and synthetic media built to deceive. Written to be read by anyone, not by compliance officers.

Current advisories →

News & action

A continuously maintained public record of what is actually happening in AI — the claims, the incidents, the money, and the policy — with the sources named so readers can check for themselves.

The record →

Approved guardrail program

We identify consumer tools that meet our technical bar for real control — and we pay for them, so that students, first responders, and families are not priced out of their own safety.

The program →

Open standards: a kill switch anyone can check

KS-1.0 · Verified Stop Standard

California has asked for AI kill switches verified by independent organizations. KS-1.0 says what verified means: five plain tests, a record anyone can check offline, and no product named. Draft 0.9.2 is open for public comment until November 15, 2026.

Read the standard and comment →

Verified Stop register

A public list of AI deployments that have passed an independent verification, who verified them, and when. Nothing is listed yet, and nobody may claim conformance until it is.

The register →

Our position: responsibility is shared, and nobody is excused

Most of the argument about AI safety is an argument about whose fault it is. We think that argument is a luxury. Responsibility sits in three places at once, and all three are currently underperforming.

1. The person

You are responsible for knowing what runs on your devices, what has your card, and what can speak in your name. That obligation is not unfair — but it is impossible without tools, and until recently no consumer tools existed. We fund those tools.

2. The provider

Companies that build and deploy AI are responsible for what their systems do after they leave the lab: for labeling, for limits, for shutdown that actually shuts down, and for not quietly turning an assistant into a salesman.

3. The government

Government is responsible for floors, not ceilings — a minimum standard of disclosure, an enforceable stop, and consequences when a synthetic voice empties an account. Floors are set slowly. That is why the other two pillars cannot wait.

Read the full framework

A note on fear. The Alliance does not trade in panic, and we do not publish predictions of the end of the world. We publish what is documented, we name our sources, and we tell people what they can do this week. Fear without a next step is just noise, and noise is how people get robbed.