BargainingFor OurMinds
An open letter from scientists
to the frontier labs
To the frontier labs,
As researchers at the frontiers of our respective fields, we have helped bootstrap your models through our published work, without receiving compensation or having meaningful input into their development.
While we want these systems to advance science and improve people’s lives, we are deeply alarmed about their catastrophic potential, and feel a responsibility to help shape their safe development.
We join the millions who use your models in celebrating their accelerating ability to enhance our work and expand our horizons. Yet we worry that the records of our work, which include unpublished research, new ideas and expert judgement, will help build increasingly powerful systems while leaving us with little say in their development or share in the value we create.
We believe this moment calls for a new labor movement. The people whose work makes these systems possible must have the collective power to shape their development and the future they are helping to create.
We therefore ask you to negotiate a collective bargaining agreement that fairly compensates us for our future contributions and gives us a continuing say in how these systems are developed and used.
As a first step, we have updated our permissions in your applications to exclude our conversations from your training data, and ask for explicit confirmation that you are honoring this choice.
James Barnes
Research Router Co-Op
Add your name
Loading sign-in options…
About the agreement
What do we contribute when we work with AI?
A scientist may work through an unpublished idea with an AI tool, then correct an answer or explain why an approach will fail. That exchange records original research and judgement developed over years of work.
As models become more capable, we can bring more demanding work into these exchanges. Permission to train on this record is something we can negotiate together. We want these contributions to improve AI and to share in the value they create.
Why bargain as a group?
Where we control training permissions, each of us can decide whether to enable them. A shared commitment lets us put common terms to the labs and choose representatives to negotiate them.
The basis for negotiation is permission to train on future research conversations and agent sessions. Together, we can offer access to records of scientific judgement on agreed terms. Scientists would retain the choice of which records to contribute.
What would the agreement cover?
The agreement would set terms for our future contributions to AI development, including permission to train on records of research conversations and agent sessions. These records can include unpublished work we bring into a session, as well as ideas we develop while using the tools.
Our first commitment concerns training permissions wherever we control them. Publication and collaboration remain governed by existing agreements and institutional obligations. Any future contribution would also depend on the rights and permissions held by the people involved.
The agreement would specify which records it covers and which uses it permits. It would also need a way to check compliance and address breaches.
How would our contributions be valued?
That is one of the things the agreement must establish. The letter asks for an appraisal method that scientists and labs can both examine and challenge.
The assessment should explain what is being valued and the evidence behind it. It should account for how contributions become useful together and inform fair compensation.
What would a seat at the table involve?
Scientists’ representatives would take part in decisions about how the labs develop and use the systems our contributions help build. Regular reviews would give them a way to examine what is happening and raise concerns that the labs must address.
The agreement would need to define that role. Scientists should be able to choose their representatives and hold them accountable.
What does signing mean?
Signing means confirming that you have updated your permissions to exclude your conversations from training data and joining the call for a collective bargaining agreement.
Review the tools you use, including any separate permissions for feedback or agent environments. If your institution controls the settings, ask its administrator to review them. The guide explains each provider's training terms and exceptions.
