A new international research fellowship is inviting technologists, AI researchers, and software security experts to tackle one of artificial intelligence’s toughest problems: ensuring machine-generated code is safe, reliable, and verifiable.
- A new international research fellowship is inviting technologists, AI researchers, and software security experts to tackle one of artificial intelligence’s...
- Apart Research, working with Atlas Computing has launched applications for the 2026 Secure Program Synthesis Fellowship, a remote mentor-led programme...
- The fellowship will focus on secure program synthesis, a field that combines artificial intelligence with formal verification methods to ensure...
- A third research track will explore spec-driven development, where multiple software implementations are generated from one specification and evaluated for...
Keep reading for the full breakdown on AI research — everything you need to know is covered below.
Apart Research, working with Atlas Computing has launched applications for the 2026 Secure Program Synthesis Fellowship, a remote mentor-led programme focused on AI safety, formal verification, and secure software development.
The fellowship will run from June to September 2026 and bring together researchers from diverse technical backgrounds to work on emerging risks associated with AI-generated software.
Fellowship Targets AI-Generated Code Risks
The programme arrives as concerns continue to grow over the rapid adoption of large language models for software development. Researchers and industry experts have increasingly warned that AI-produced code can contain hidden flaws, security gaps, or unintended behaviour if not properly tested and verified.
The fellowship will focus on secure program synthesis, a field that combines artificial intelligence with formal verification methods to ensure software behaves according to defined requirements. Participants will work in small research teams alongside mentors and project managers.
Organisers say the initiative aims to strengthen trust in AI-assisted coding systems by improving how software specifications are written, validated, and tested.
Four Major Research Areas Announced
The programme is structured around four research themes.
The first focuses on specification elicitation, in which researchers will study methods for converting informal instructions and software documentation into structured, formal models. Another stream will examine specification validation, testing whether those formal specifications accurately reflect intended system behaviour.
A third research track will explore spec-driven development, where multiple software implementations are generated from one specification and evaluated for correctness and reliability.
The final research area centres on adversarial robustness. Participants will investigate how malicious inputs or unexpected prompts can disrupt AI reasoning systems and automated verification pipelines.
Remote Format with Flexible Participation
The fellowship will operate entirely online and is expected to require between 8 and 30 hours per week from participants. Teams will include mentors, research fellows, and project management support from Apart Research.
Participants will also gain access to compute resources, API credits, research guidance, and opportunities to present their work during a final demo day. Selected projects may also receive support for conference participation and awards for outstanding contributions.
Applications for participants close on 31 May 2026, while selected fellows are expected to be announced on 9 June. Research activities will begin on 15 June.
Broad Technical Backgrounds Encouraged
The organisers are encouraging applications from candidates with experience in areas such as formal proof engineering, secure systems, reverse engineering, theorem proving, AI evaluation, and software testing.
Unlike many research programmes, the fellowship does not require a fixed academic pathway, opening the door to applicants from a wide range of technical and interdisciplinary fields.
The initiative reflects growing global interest in AI safety research as governments, universities, and technology organisations seek stronger safeguards for increasingly autonomous software systems.