This matters especially in frontier AI, where the exact shape and form of risk arising may be difficult to predict.

Regulators and the public should hence have a strong interest in empowering those who are able to spot risk first when and where they arise.

Former and current lab employees have called for such protections. 

This is why we work to strengthen the rights of insiders at the frontier of AI. We do this by providing feedback on proposed regulation, often in collaboration with other civil society organizations, and especially Kosmas Zittel of the Whistleblower Netzwerk e.V.

Europe has taken the globally leading role in implementing whistleblowing protections specific to risks from AI. Concretely, the EU AI Act will be covered under the scope of the Whistleblowing Directive from August 2026, with potentially more protections resulting from the Code of Practice already at an earlier date.

Outside of direct benefits to individuals covered by the Directive, primarily individuals in Europe, this means Europe is setting global precedent. This ‘spotlight’ on the the effectiveness of its protections has the potential to shape protections globally. It therefore matters that protections are implemented well and lessons are distilled for regulators globally

Together with our partner organization Whistleblower Netzwerk e.V., who have almost 2 decades of experience in policy work in Europe, we believe in 4 focus areas as it relates to EU Whistleblowing Protections:

1. Establishment of a well-resourced EU AI Office Whistleblower Mailbox
2. Whistleblower Directive Roll-out
3. Sharing learnings with policy developers globally

While we recognize the limitations of the Code of Practice inherent to its format, we nonetheless believe in its potential. We have lead feedback efforts on the Code of Practice Second and especially Third Draft to ensure that non-retaliation commitments made by labs are extensive and can be relied upon.

Find our feedback on the second draft here.

If you are interested in our feedback on the Third Draft, please reach out to us directly

Legal protections alone are not sufficient. We believe a well-equipped Whistleblower Mailbox is a critical element to ensure that reports are investigated and handled appropriately. This allows individuals willing to speak up to receive the support they need and ensures that lab-internal whistleblowing policies deliver benefits for both labs, insiders, and the public. It also sets further precedent on how to establish great infrastructure for AI whistleblowing globally.

We have historically and are continuing to provide input on how to design a strong Whistleblower Mailbox to relevant recipients.

If you are interested in our thinking and documents outlining our proposals, please reach out to us directly

By virtually all measures, the U.S. is the global leader in frontier AI development. We estimate approximately 15,000 individuals work at the AI frontier in the U.S., with about 5,000 of those specifically in frontier research and engineering.

While arguments can be made that some whistleblower protections already exist today that might cover risks from frontier AI, such as under the SEC program (contact us if you’d like more information), none of the employees mentioned above currently have legal clarity regarding when and where they can benefit from whistleblower protections when reporting concerns about serious risks from AI.

This situation is problematic.

We have previously and are currently providing feedback on policy proposals in California and U.S. federal level.

As examples:

1. A paper we contributed to, published with the Center for AI Policy
2. Our thoughts on SB 53
3. Our thoughts on the Joint California Working Group’s Recommendations

We collaborate with civil society organziations to provide non-partisan input on policy proposals and help set agendas.

We are always happy to work together to further our mission. With our without public attribution. We only care about the impact.