AI whistleblower protection guide
If you’ve seen something at an AI company that crosses a line — a suppressed safety eval, a privacy breach, biased outputs shipped anyway, or deceptive claims to investors — speaking up is daunting and the rules are confusing. This guide explains the protections that apply where you are, how to preserve evidence, and which channels keep you safest, so you can make an informed decision rather than a fearful one.
How it works
You select your jurisdiction (UK, EU, US, or other) and the type of concern. The tool surfaces the relevant laws and the protections they give — for example the UK’s PIDA, the EU Whistleblower Directive and AI Act reporting channels, or US sector statutes — and explains how those protections work in practice. It then provides concern-specific steps for documenting and routing the issue, general guidance on internal-versus-external and anonymous reporting, and a list of regulators matched to the concern (data-protection authorities, financial regulators, safety bodies).
Tips and notes
- Protect yourself first. Research and contact advisers from personal devices and accounts, never employer-monitored systems.
- Internal first, usually. It typically keeps protection strongest and builds a record — unless internal channels are compromised or harm is severe.
- Preserve, don’t exfiltrate. Keep evidence of the concern, but don’t remove originals or copy personal data beyond what’s strictly necessary.
- Get advice before going external. A whistleblowing charity or employment lawyer can be the difference between protected and unprotected disclosure. This guide is educational, not legal advice.