Sprint projects

Noticing the escapes: Who spots a runaway agent, and how long does it take?

The Escapees

Strengthening International AI Incident Discovery

Team A-O-I

When to Ask: An RL Environment That Teaches AI Agents to Act on What Users Mean and Not Just What They Say

Teachafy

Declare Before You Run: An Open Filing Standard for Frontier Training and Evaluation Runs

TeamFourStar

When the Evaluation Is the Incident: Testing AI Incident-Reporting Regimes on the OpenAI–Hugging Face Intrusion

Arathi

Readable but Not Causal: Limits of Self-Attributed Welfare Representations in Language Models

Project Anchored

Team Wagner

Model, Instance, or Persona? Measuring Affective Signals in Public Text After an AI Is Retired

One Dial, Not a Tree: Occupational Personas and Emergent Misalignment

Misbehaved

Sisyphus in the loop: What Makes an LLM Persist?

Scaling Sisyphus

Omission Attacks: When Doing Nothing Is the Attack

MAIA

Detecting LLM Subversion in Vulnerability Patching Settings

Vuln4Control

ActionLens: Pre-Execution Environment Probing for Agent Action Approval

Udbhav&Ashok

Undoing the Fix: CVE Regression as an AI Control Setting

Counterfactual Components for AI Control Protocols

Plausible Counterfactual Guys

Syscon: Kernel-Level Syscall Monitoring for AI Agent Control in Sandboxed Environments

Syscon

LidaSim: Testing AI Policies With Persona-Based Simulations

Lida Safety

Markov Chain Lock Watermarking: Provably Secure Authentication for LLM Outputs

MCL

Political Intelligence for AI Safety: The AI Risk Attitudes Survey (AIRAS)

AIRAS

Fingerprinting All AI Cluster I/O Without Mutually Trusted Processors

The Bypass Annihilators

Prototyping an Embedded Off-Switch for AI Compute

Who Does Your AI Serve? Manipulation By and Of AI Assistants

Cart Abandonment Issues 🛒

Eliciting Deception on Generative Search Engines

Ardy

Cross-Linguistic Sycophancy in Frontier LLMs: A Benchmark Study

Talex