Resources
White papers, datasets, benchmarks, our podcast, and the latest insights — everything we're publishing on aviation AI.
Never miss an episode
Get new Travel Tech Podcast episodes, aviation AI insights, and use-case deep dives delivered to your inbox.
Nobody Would Fund a Human for This
Nobody would fund a human team to call every hotel and confirm every booking — and that precise economics gap is exactly what makes AI the right tool for the job. Mike Putnam, Anne Marie Pellerin, Yagub Rahimov, and Fred Bean are four founders who reached the same insight from opposite corners of travel and aviation, and host Alex Brooker lets each of them make the case in their own words.
Listen & browse all episodes
Pre-Flight Benchmark Dashboard
A live dashboard of how leading AI models perform on aviation safety tasks. Compare frontier hosted models with open-weight local ones, ranked against a human-expert baseline. We update it as we run new evaluations.
View Dashboard
Prompt Injection Risk in Aviation AI: White Paper
1,776 adversarial test cases against commercial AI systems processing standard aviation data formats. 43% of attacks produced incorrect safety-critical outputs. Read the full methodology, results, and mitigation framework.
Learn More
Aviation AI Use Case Directory
Browse our directory of aviation AI applications, from safety systems to customer service. See how AI is being used across aviation operations.
Learn More
Airside Labs on Hugging Face
Browse our open-source aviation AI datasets on Hugging Face — including the Pre-Flight benchmark (300+ questions) plus NOTAM, ADS-B and aviation NLP datasets for training and evaluating aviation AI systems.
Visit Resource
Pre-Flight Benchmark on UK AISI Inspect AI
The Pre-Flight aviation intelligence benchmark is now available on the UK AI Security Institute's Inspect AI evaluation framework. Evaluate AI systems against aviation-specific criteria.
Visit ResourceLatest Insights
Explore our latest research, analysis, and industry updates
Top Aviation AI Startups Building the Future of Flight in 2026
An honest, sourced map of the startups building aviation AI in 2026 — flight operations, cockpit copilots, autonomy, and the truth about aviation foundation models — plus the independent assurance layer that proves these systems are safe.
Top Platforms for Evaluating Aviation AI in 2026
A field guide to the platforms and benchmarks for evaluating aviation AI in 2026 — domain-specific benchmarks, general eval platforms, security firms, and consultancies — and how to choose the right one for your use case.
Abundance of Competence, Scarcity of Trust
The UK government's AI Scenarios 2030 models how AI's gains get divided, yet not how cheap competence grows the pie, or why, when everything starts to look good, trust and permission become the only scarce goods.
Prompt Injection Risk in Aviation
1,776 adversarial test cases against LLMs processing standard aviation data formats reveal systematic vulnerabilities in how large language models handle prompt injection in safety-critical contexts.
98.7% Retrieval Accuracy from Metadata You Already Have
Fine-tuning embedding models for aviation NOTAM retrieval — and why bigger models aren't always better out of the box.
Comparative Analysis: Pre-Flight vs MITRE/FAA ALUE Benchmarks
A comprehensive analysis of two pioneering aviation LLM assurance benchmarks, examining how Airside Labs' Pre-Flight and MITRE/FAA's ALUE address distinct operational layers in aerospace AI safety.
Alternatives to Big Cyber for LLM Pen Testing
When organisations think about AI security testing, many automatically turn to established cybersecurity firms. But LLM penetration testing requires fundamentally different expertise.
Customer AI Chatbot Flying Blind: The Hidden Risks
A comprehensive analysis of 11 leading language models reveals critical safety gaps that could ground your customer service operations.
Crescendo: How Escalating Conversations Break AI Guardrails
Why single prompt testing misses the most dangerous AI failures and how the crescendo technique is exposing critical vulnerabilities in customer service systems.
Alternative to Big Four AI Testing: Why Domains Matter
The AI revolution is sweeping across industries faster than ever, but when it comes to testing and validating these AI systems, many organisations are turning to generic frameworks.
Airside Labs Responds to the UK AI Opportunities Action Plan
At Airside Labs, we're committed to advancing aviation technology through innovative AI solutions while maintaining the industry's paramount focus on safety.
Airside Labs Responds to UK CAA's AI in Aerospace Request
At Airside Labs, we're committed to advancing aviation technology through innovative AI solutions while maintaining the industry's paramount focus on safety.
Airside Pre-Flight Benchmark Joins AISI Evaluations Package
Aviation AI Benchmark Now Available Through UK's AI Security Institute's inspect_evals Framework
PRESS RELEASE: Airside Labs Launches Pre-Flight AI Benchmark on GitHub
Aviation AI Testing Framework Now Available for Industry Contributions
Aviation Eval – Flight Test 1: Anthropic Models Compared
With the exciting release of Anthropic's updated Sonnet and Haiku models, we're sharing our first evaluation results from the Pre-Flight benchmark.
Airside Labs Launches at Royal Aeronautical Society Event
New Aviation AI Venture Unveils Industry-First Benchmark for AI Model Testing
Preflight Aviation Intelligence Benchmark: Contributor Guide
We are collecting the most challenging and comprehensive set of aviation-related questions ever assembled for AI evaluation.
GAIA: Benchmarking Next-Gen AI Assistants for Aviation
Benchmarks play a crucial role in measuring AI progress and setting new standards. GAIA is one such benchmark that has caught our attention.
Open-Source Aviation AI Datasets
Airside Labs publishes open-source aviation datasets on Hugging Face for the community to train and evaluate aviation AI. Direct links below, or browse the full organisation at huggingface.co/AirsideLabs.
300+ multiple-choice questions testing AI on ICAO standards, airport ground operations, flight dispatch, and emergency procedures — the dataset behind our arXiv paper (arXiv:2607.01829) and UK AISI Inspect AI integration.
A dataset of NOTAMs (Notices to Air Missions) for training and evaluating aviation natural-language models on operational notices.
A NOTAM corpus for aviation language, classification, and information-extraction tasks.
Automatic Dependent Surveillance–Broadcast flight-surveillance data for aviation AI and trajectory analysis.
Help and documentation content for aviation portal support, useful for retrieval and question-answering tasks.
What does Airside Labs do?
Airside Labs builds and independently tests AI systems for aviation. We build AI prototypes, fine-tuned embedding and language models, RAG systems and data-analytics assistants, and we develop domain-specific benchmarks, adversarial datasets, and assurance tooling. The Pre-Flight benchmark is one tool among many. It is integrated into the UK AI Security Institute's Inspect AI framework. What makes the systems better is use case specific testing and red teaming. We map findings to the EU AI Act, OWASP LLM Top 10, NIST AI RMF, and MITRE ATLAS for audit ready evidence.
Generic "Big Four" consultancy and cybersecurity firm AI testing covers governance breadth but not aviation's safety critical context, where the same model failure is a minor annoyance in one industry and a safety hazard in another. For National Airspace System integration, the MITRE/FAA ALUE benchmark is complementary. See our Pre-Flight vs ALUE comparison or the full aviation AI evaluation & benchmarking landscape.