Amazon Nova AI Challenge returns with Nova Forge access for competing teams

2026-02-02 · Source: Amazon Science homepage · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cybersecurity & Data Privacy · Depth: Advanced, medium

Summary

The Amazon Nova AI Challenge returns for its second year, bringing together ten university teams from five countries to build trusted AI software agents. This year, the competition shifts focus from single-task code vulnerability identification to multi-step software development projects, requiring AI agents to plan, build, and test changes across entire codebases while ensuring utility and safety. A key new feature is the integration of Amazon Nova Forge, providing academic teams with unprecedented access to Nova models, custom training environments, model compression tools, and safety controls. This allows teams to customize Nova models into "Novellas" by adding their own data and research methods throughout the training process. Teams receive $250,000 in sponsorship and AWS credits, competing for prizes up to $250,000, with final evaluations and winner announcements scheduled for October 2026.

Key takeaway

For AI Scientists and Research Scientists developing secure AI agents, the Amazon Nova AI Challenge highlights the critical need to integrate safety and utility from the ground up. Your work should prioritize building agents capable of multi-step software development while simultaneously implementing robust security measures and leveraging advanced customization platforms like Nova Forge to enhance model trustworthiness and performance.

Key insights

Academic teams can now customize frontier AI models using Nova Forge for secure, multi-step software development.

Principles

• AI agents require dual focus on utility and safety.
• Red teaming drives adaptive security improvements.
• Access to model customization accelerates academic research.

Method

Teams use Nova Forge to access Nova model checkpoints, add custom data during training, and create "Novella" models with specific secure coding patterns, training environments, or agentic security reasoning.

In practice

• Utilize custom training environments for real-world scenarios.
• Apply model compression for cost-effective performance.
• Implement built-in safety controls for secure AI agents.

Topics

• AI Agents
• Nova Forge
• Secure Software Development
• Model Customization
• Generative AI

Best for: AI Scientist, Research Scientist, AI Student, AI Researcher, Software Engineer

Related on AIssential

• Engaging the AI community through building, research, and shared learning — Community engagement through hackathons and academic challenges accelerates AI innovation and addresses real-world problems.
• Automating competitive price intelligence with Amazon Nova Act — Amazon Nova Act enables robust, scalable browser automation for competitive intelligence using natural language and programmatic logic.
• The Race to Put AI Agents Everywhere — AI agents are now viable, driving a rapid "clawification" of software and a sprint to enterprise productization.
• Fable 5 IS BACK? GPT 5.6 & Gemini 3.5 Pro Delayed, NEW OpenAI AI Chip, & QWEN STEALING! AI NEWS — AI model capabilities can pose national security risks and drive industrial-scale espionage.
• Amazon Bedrock AgentCore vs Reality — Building reliable AI agents is primarily a software engineering challenge, not just an AI/prompting one.
• This AI Startup’s Army Of 15,000 Hackers Pressure Test Claude, GPT-5 And Gemini - Forbes — Human red-teaming data is crucial for training AI agents to proactively identify and mitigate complex AI system vulnerabilities.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Amazon Science homepage.