CALL FOR APPLICATIONS 2026: Secure Program Synthesis Fellowship on AI Security and Formal Methods

Apart Research has announced the launch of the Secure Program Synthesis Fellowship, a new remote research programme exploring the intersection of formal methods, AI systems, software correctness, and cybersecurity. The initiative, developed in partnership with Atlas Computing, will run from June to September 2026 and aims to support collaborative research projects tackling some of the most pressing challenges in AI-generated code and secure software development.

As artificial intelligence continues to accelerate software production, concerns around software correctness, verification, and security are becoming increasingly significant. The fellowship is designed to explore how advanced verification techniques and formal methods can improve confidence in AI-generated code, particularly in systems where errors or vulnerabilities can have major consequences.

The programme will bring together mentors, project managers, and research participants to work on underspecified and technically challenging problems related to specification, validation, and adversarial robustness in AI-supported software systems.

Focus Areas of the Fellowship

The Secure Program Synthesis Fellowship will support research projects across several specialized focus areas connected to formal verification and AI safety.

Key research themes include:

• Specification elicitation and formalization.
• Validation of extracted software specifications.
• Spec-driven software development and evaluation.
• Adversarial robustness for formal methods and quality assurance tools.
• AI-assisted verification systems.
• Security and correctness in AI-generated code.

Participants will work on mentor-led projects exploring how formal specifications can be extracted from ambiguous or incomplete sources such as legacy systems, documentation, or human stakeholder requirements.

The programme also emphasizes improving methods for verifying whether software specifications accurately capture intended system behavior, especially in increasingly automated development environments.

Why Secure Program Synthesis Matters

The fellowship is based on the idea that as AI-generated code becomes cheaper and faster to produce, the main challenge shifts from implementation to validation and specification.

According to the programme description, many real-world systems lack complete or clearly defined specifications, making it difficult to verify whether generated code behaves as intended. Weaknesses at the specification level can lead to errors and vulnerabilities that propagate throughout downstream implementations.

The fellowship therefore focuses on improving methods for:

• Eliciting accurate specifications.
• Formalizing system requirements.
• Validating intended software behavior.
• Identifying weaknesses in automated reasoning systems.
• Strengthening adversarial robustness in formal verification pipelines.

The initiative reflects growing interest within the AI safety and cybersecurity communities in developing scalable approaches to software verification and oversight for increasingly autonomous coding systems.

Programme Structure and Timeline

The fellowship will run for approximately three to four months, beginning in June 2026 and concluding with final submissions and a demo day later in the year.

Important programme dates include:

• Mentor applications: 28 April – 9 May 2026.
• Participant applications open: 19 May 2026.
• Participant applications close: 31 May 2026.
• Participant announcements: 9 June 2026.
• Projects begin: 15 June 2026.
• Mid-project presentations: July and August 2026.
• Final submissions: August and September 2026.
• Demo day: September or October 2026.

The fellowship is entirely remote and designed to accommodate participants dedicating between 8 and 30 hours per week.

Collaborative Research Teams

Each research team will include:

• An expert mentor.
• An Apart Research Project Manager.
• Research participants and mentees.

The programme emphasizes rapid experimentation, collaborative problem-solving, and the creation of meaningful intermediate research outputs even when problems remain difficult or partially unsolved.

Participants will also receive support including:

• Research guidance.
• Compute resources.
• API credits.
• Access to research communities.
• Potential conference or demo-day travel support.
• Prize opportunities.

Mentors and Technical Expertise

The fellowship will feature mentors and contributors from the fields of formal methods, software verification, and AI security.

The programme specifically encourages applications from individuals with backgrounds in:

• Proof engineering.
• Secure systems design.
• SMT solving and model checking.
• Red teaming and penetration testing.
• Reverse engineering and fuzzing.
• AI agent development.
• Machine learning evaluation systems.
• Formal verification tooling.

However, organizers also note that no specific academic or professional background is required, encouraging applicants from diverse technical disciplines to participate.

Flexible Participation and Research Opportunities

The fellowship is intended to support exploratory and high-impact research collaborations that can potentially lead to workshop papers, conference submissions, demonstrations, or publishable research outputs.

Participants are allowed to join while employed full-time, provided they can dedicate at least eight hours per week to the programme.

Although the fellowship is generally unpaid, organizers noted that stipends may be available in situations where compensation would enable participation.

The initiative reflects growing international attention toward AI safety, software assurance, and formal verification as organizations seek ways to ensure increasingly autonomous systems remain secure, reliable, and aligned with intended behavior.

Interested applicants can learn more through the official Apart Research website or contact the organizers through secure-program-synthesis-fellowship@apartresearch.com.

LEARN MORE

Join Our WhatsApp Community

Disclaimer: Remote Job Opportunities (RJO) is not the organization offering this opportunity. For any inquiries, please contact the responsible organization directly. Please do not send your applications to RJO, as we are unable to process them. Due to the high volume of  emails we receive daily, we may not be able to respond to all inquiries. Thank you for your understanding.