Applications are now open for the Frontier AI Security Training (FAST) 2026, a fully funded five-day intensive programme taking place in Singapore from 28 September to 2 October 2026. Designed specifically for experienced cybersecurity and machine learning practitioners, FAST equips participants with the practical skills needed to defend and secure frontier artificial intelligence (AI) systems against emerging threats.

As AI technologies continue to evolve rapidly, new vulnerabilities and attack surfaces are emerging faster than conventional cybersecurity approaches can address them. The FAST programme bridges the gap between AI and cybersecurity by bringing together a highly selective cohort of professionals from around the world to explore the latest techniques in AI security through lectures, practical labs, collaborative exercises, and one-on-one mentorship.

Participants will work alongside internationally recognized instructors, researchers, and fellow practitioners while gaining hands-on experience attacking and defending frontier AI models, autonomous agents, and AI-powered systems.

About the Frontier AI Security Training (FAST)

Frontier AI Security Training (FAST) is a highly specialized technical programme designed to accelerate the development of professionals capable of securing the next generation of advanced AI systems.

The programme focuses on practical AI security rather than theory alone. Participants spend five intensive days exploring the latest research while completing hands-on exercises involving real-world AI security challenges.

FAST addresses one of today’s fastest-growing technology challenges: ensuring that increasingly capable AI systems remain secure, reliable, and trustworthy.

The programme is supported through philanthropic funding, making participation free for selected applicants.

Why Frontier AI Security Matters

Artificial intelligence is advancing at an unprecedented pace.

Modern frontier AI models and autonomous agents are creating entirely new cybersecurity challenges that traditional security practices were never designed to handle.

These challenges include:

• Prompt injection attacks
• AI model manipulation
• Model backdoors
• Data poisoning
• AI system verification
• Autonomous agent security
• AI alignment
• Monitoring increasingly capable AI systems
• Security assurance for frontier models

Because the field is evolving so quickly, there is currently no universally accepted playbook for defending advanced AI systems.

FAST was created to help build a new generation of experts capable of protecting these emerging technologies.

Why You Should Apply

FAST offers a unique opportunity to gain practical experience in one of the fastest-growing fields in technology.

Participants will:

• Learn directly from internationally recognized AI security experts.
• Develop practical experience through hands-on laboratory exercises.
• Attack and defend frontier AI models.
• Collaborate with experienced professionals from around the world.
• Gain exposure to current AI security research.
• Build long-term professional relationships.
• Explore career pathways in AI security.
• Receive guidance on future research opportunities.
• Join a growing international AI security network.

Programme Details

Programme Format

• Five-day intensive in-person programme
• Fully funded
• Technical training track
• Highly interactive laboratory sessions
• Small cohort learning

Programme Dates

28 September – 2 October 2026

Location

Singapore

Cohort Size

Only 20 participants will be selected.

The small cohort allows for:

• Individual attention
• Close collaboration
• Personalized mentoring
• Interactive discussions
• Team-based laboratory work

Who Should Apply?

FAST is designed for experienced technical professionals rather than beginners.

The programme primarily targets two groups.

1. Security Engineers

Ideal applicants include professionals with practical experience in:

• Security engineering
• Offensive security
• Infrastructure security
• Vulnerability research
• Reverse engineering
• Penetration testing

Applicants whose experience is mainly in compliance, auditing, or advisory roles may not find the programme suitable.

2. Machine Learning Engineers and Researchers

Applicants should have practical experience with:

• Training machine learning models
• Fine-tuning large language models
• PyTorch
• Deep learning
• Neural networks
• Transformer models

Participants should already have hands-on experience running model training rather than only theoretical knowledge.

Recommended Background Knowledge

Applicants are encouraged to be familiar with frontier AI research, including topics such as:

• AI alignment
• Global workspace language models
• High-stakes AI verification
• Alignment faking
• Sleeper agents
• AI safety
• Adversarial machine learning

Applicants who are unsure whether they meet the experience requirements are still encouraged to apply, as the programme places greater emphasis on demonstrated technical ability than formal credentials.

What Participants Will Learn

The programme combines lectures, discussions, and extensive laboratory work.

Day 0 – Pre-Programme Preparation

Participants will:

• Prepare their technical environment.
• Review prerequisite materials.
• Refresh AI fundamentals.
• Ensure everyone begins with a common technical foundation.

Day 1 – AI Model Fundamentals

Topics include:

• Model behavior
• Prompt injection
• Log probabilities
• Instruction hierarchies
• AI guardrails
• Failure modes

Day 2 – AI Control

Participants will explore:

• Monitoring systems
• AI supervision
• Control protocols
• Building defences for untrusted AI systems

Day 3 – Open-Weight Model Security

Hands-on exercises include:

• Removing safety fine-tuning
• Data poisoning
• Introducing AI backdoors
• Model distillation
• API-based attacks

Day 4 – Verification and Emerging Research

Participants will study:

• Compute verification
• AI assurance
• Security verification
• Open research challenges
• Hardware verification

Day 5 – Career Development

The final day focuses on:

• Research project planning
• Instructor feedback
• Career pathways
• Fellowships
• Grants
• Research opportunities
• Funding opportunities
• One-on-one mentoring
• Professional networking

What Participants Will Gain

By the end of the programme, participants will leave with:

• Practical AI security skills
• Hands-on laboratory experience
• A clearly defined next career step
• Connections to AI researchers and practitioners
• Funding and hiring insights
• Long-term alumni support
• A global professional network

Learning Environment

FAST is designed around practical learning.

Participants can expect:

• Short morning lectures
• Extensive laboratory work
• Instructor mentoring
• Pair programming
• Small group discussions
• Hands-on experimentation

No software installation is required.

The programme provides:

• Computing resources
• Technical environment
• Workspace
• Food during the programme

World-Class Instructors

Participants will learn from internationally respected AI security experts.

The instructional team includes professionals affiliated with organizations such as:

• ERA
• Niantic
• Mila
• University of Oxford
• University of Cambridge
• UC Berkeley
• IBM
• Microsoft
• Meta
• LinkedIn
• GovTech Singapore
• UK AI Security Institute

Lead instructors include:

• David Williams-King
• Jannis Kirschner
• Bary Levy
• Nitzan Shulman

Together, they bring extensive expertise in:

• AI alignment
• AI safety
• Offensive cybersecurity
• Reverse engineering
• Systems security
• Adversarial machine learning
• Secure software engineering
• AI security research

Funding and Financial Support

The Frontier AI Security Training programme is fully funded through philanthropic grants.

Selected participants receive:

• No tuition fees
• Free participation
• Workspace throughout the programme
• Meals during the five-day training
• Computing resources
• Learning materials

Additionally:

• Travel support is available based on financial need.
• Accommodation support is also provided on a needs basis.

The organizers emphasize that financial limitations should not prevent qualified candidates from participating.

Selection Process

Admission is competitive due to the limited cohort size.

Applicants will be evaluated based on:

• Technical experience
• Practical skills
• Demonstrated ability
• Relevant professional background
• Potential contribution to the AI security field

Only 20 participants will be admitted.

Career Opportunities After FAST

Graduates gain access to:

• AI security research opportunities
• Fellowships
• Grant opportunities
• AI safety organizations
• Frontier AI research labs
• Professional mentorship
• International alumni network
• Career guidance
• Research collaborations

How to Apply

Interested applicants should submit their application before the application deadline through the official Frontier AI Security Training application portal.

Because places are extremely limited, early application is strongly encouraged.

Apply Here

Deadline

Application Deadline: 6 August 2026

For more information, visit the official website.

Discover more global training opportunities on OFY, VISIT HERE