Agent Safety Infrastructure
Build the Airlock and reusable infrastructure connecting proof checks to tools, resources and execution.
Bocconi · REALMilan, ItalyCall upcomingRole details
Research and engineering
Get involved
Help develop the methods and infrastructure behind Proof Carrying Contracts. The programme brings together research in formal methods, machine learning and agent systems, with opportunities to lead a distinct contribution through an appointment or take part in scientific discussion and external evaluation.
Open positions
7 planned appointmentsThe following research and engineering roles are being prepared. Each has a primary base in Milan or Turin and collaboration across the participating groups. Application instructions will appear with each call as it is finalised.
Postdoctoral research
Contract Calculi and Verifiable Agent Tools
Develop formal foundations and reusable representations of domains, resources and tool effects.
AI4I · PRAXISTurin, ItalyCall upcomingRole details
Contract Calculi and Verifiable Agent Tools
Connect contract semantics and compositional guarantees to verifiable action and delegation.
Bocconi · REALMilan, ItalyCall upcomingRole details
AI and Formal Methods for Intent Contractualization
Study how agents capture ambiguous requirements, formulate contracts and expose unintended permissions.
AI4I · PRAXISTurin, ItalyCall upcomingRole details
Learning and Evaluation of Safe Agentic Models
Develop and evaluate capable agents that plan and act under explicit constraints checked by the system.
Bocconi · REALMilan, ItalyCall upcomingRole details
Multiagent Coordination and Verifiable Delegation
Investigate how interaction structures and delegation affect commitments across different principals.
AI4I · PRAXISTurin, ItalyCall upcomingRole details
The role pages explain each research direction, its contribution to the programme, the working environment and the appointment details.
Beyond an appointment
Collaboration and community
Scientific discussion, external scrutiny and contributions from other implementations will help develop the protocol. Researchers and developers are invited to bring relevant methods, difficult requirements and concrete interaction scenarios.
Follow the programme
Follow the foundations paper, implementation releases and opportunities for public discussion.
Release announcements will appear on this website. A dedicated updates channel is being prepared.
Discuss and collaborate
Bring a research connection, a requirement that is difficult to express, or an interaction in which agents need to rely on a commitment. Relevant discussions include contract semantics, tool interfaces, delegation and adversarial behaviour.
A project contact route will be added before public launch.
Test and contribute
Upcoming with the first implementation release
The first implementation release will provide examples to run, contracts and tool cases to examine, and an implementation to test under adversarial conditions. Supported contribution tasks and instructions will be published with the relevant releases.
Discussion of the paper and its assumptions is welcome before those implementation resources become available.
Contribution instructions forthcoming