Publications

Signed scholarly work

1. AI Safety 2026: From Risk to Resilience
   Owen Sakawa; Bitange Ndemo; Moussa Bello; Aisha K. Patel; Marcus A. Rodriguez; Yan Zhu; Mina Narayanan; Suhani Gharial; Daniela Muhaj; Bosco Hung; Victoria Snorovikhina; Alexander Saeri; Jess Graham; Michael Noetel; Neil Thompson · 2026-04-03
   This report synthesizes the sharpest signals from the 2026 International AI Safety Report and adjacent public evidence, with a focus on capability acceleration, evaluation evasion, cyber risk, and the current state of governance. PDF
2. PatternBench: Evaluating Long-Context Memory, Safety, Hallucination, and Governance in Large Language Models
   Owen Sakawa; Jackson Mwaniki; Bitange Ndemo; Randi C. Martin; Valentin Dragoi; Caleb Kemere; Krishna V. Palem; Douglas Natelson; Fernanda Morales-Calva; Stephanie Leal · 2026-02-04
   PatternBench is a benchmark paper on long-context reliability in large language models, focused on multi-turn conversations in regulated environments. The benchmark is designed to test how recall, hallucination dynamics, calibration, and governance compliance change as dialogue length increases. PDF
3. CASCADE: Cross-Agent Supply Chain Audit and Defense Ecosystem
   Owen Sakawa · 2026-02-22
   CASCADE is a framework paper on agentic AI supply-chain risk. It models autonomous systems as dependency networks so teams can identify concentration, score attack-surface exposure, and reason about how upstream failures cascade downstream. PDF
4. Aegis: A Comprehensive Framework for Continuous Security Assessment of Autonomous AI Agents
   Sarah J. Chen; Marcus A. Rodriguez; Aisha K. Patel; James R. Thompson; Owen Sakawa; Jackson Mwaniki; Leon Derczynski; Erick Galinkin · 2026-02-20
   Aegis is a multi-author paper on continuous security assessment for autonomous AI agents. It argues that stateful, goal-directed, tool-using systems create attack surfaces that are not captured by single-turn red teaming or stateless safety checks. PDF
5. Immune System for AI: A Governance Infrastructure for Responsible AI Deployment
   Owen Sakawa; Jackson Mwaniki; Sarah J. Chen; Marcus A. Rodriguez; Aisha K. Patel; James R. Thompson; Leon Derczynski; Erick Galinkin · 2026-02-20
   Immune System for AI presents a governance infrastructure for responsible AI deployment, focused on continuous security assessment, policy-aware monitoring, and intervention around autonomous systems. PDF

Lab publications

1. TruthChecker
   Elloe AI · 2025-08-08
   TruthChecker is a technical brief on claim verification for AI systems used in higher-risk environments. It is framed as a workflow for identifying unsupported claims, matching them against source material or known patterns, and creating a review trail before outputs are relied on. PDF

Authored writing

1. AI is breaking the corporate apprenticeship model
   Essay · 2026-04-20
   AI will not only change white-collar employment. It may also interrupt the hidden apprenticeship system through which firms turn early-career workers into reviewers, managers, and decision-makers. That is why the labor story is also a governance story.
2. What happens when AI is wrong after deployment?
   Insight · 2026-04-18
   The hardest AI failures rarely look dramatic at launch. They begin as small deviations in production, then become governance problems because organizations fail to detect, escalate, or stop them in time.
3. AI governance after launch
   Insight · 2026-04-18
   The launch decision is not the end of AI governance. It is the point where governance has to become operational and prove it can still control the system once conditions change.
4. Why AI compliance fails when it stops at documentation
   Insight · 2026-04-18
   AI compliance becomes fragile when it remains a paperwork exercise. Real compliance depends on whether teams can observe, review, and constrain AI behavior after deployment.
5. The EU AI Act will expose weak post-deployment controls
   Policy insight · 2026-04-18
   The most important compliance question is no longer whether organizations have a policy. It is whether they can show how live systems are governed when risk changes.
6. AI safety vs AI governance vs AI compliance
   Explainer · 2026-04-18
   These terms are often used interchangeably. They should not be. But they do converge once organizations have to govern AI systems in production.
7. How to build an AI incident response function
   Operational guide · 2026-04-18
   AI incident response should not be improvised the first time a live system crosses a threshold. It needs structure, ownership, and decision rights before the failure arrives.
8. Why I Started Elloe AI
   Founder essay · 2026
   A founder essay on purpose, timing, and why trust became the thesis behind Elloe AI.
9. Lessons from Building at 14
   Personal essay · 2026
   A personal essay on the lessons, instincts, and early building experience that shaped later work.

Related pages

• Research
• Insights
• Media and press